CHANGELOG.md

Changelog (Auto Generated by Claude Code with GLM model)

Click to expand v1.24.0 details

[1.24.0] - 2026-09-06

Added

• Stata Discovery: Add uvx stata-mcp discover to list installed Stata paths on macOS and Windows, sorted and deduplicated with one path per line; return .app bundles on macOS and .exe files on Windows without launching Stata or changing configuration
• Discovery Errors: Add OSNotSupported; unsupported operating systems print only OS not supported to stderr and exit with code 1, while no installations produces empty output and exit code 0
• Audit Trail: Add append-only per-tool JSONL lifecycle ledgers, readable run IDs, client/protocol metadata, and cross-ledger security event IDs
• Do-file Evidence: Execute immutable full-SHA-256 content-addressed snapshots and record original path, hash, size, reuse state, and generated logs
• Local Debug Tracing: Add default-on local OpenTelemetry spans and immediate checkpoints for get_data_info and stata_do, with bounded file rotation and no default network export
• Slow-call Diagnostics: Capture privacy-safe Python thread locations after 30 and 120 seconds and link slow events to Audit run IDs, trace/span IDs, and process IDs
• Cross-platform CI: Exercise modern and legacy stdio protocols on Linux, macOS, and Windows with Python 3.11–3.13
• Repository Audit Skill: Add the stata-audit skill to the repository's Stata toolbox plugin for inspecting and validating audit records, reviewing security decisions, and generating local HTML reports; the skill is separate from the PyPI package
• Audit Documentation: Add English and Chinese guides for audit events, evidence inspection, snapshot verification, security linkage, and correlation with debug traces

Changed

• MCP SDK: Migrate the high-level server from MCP Python SDK 1.x FastMCP to stable 2.x MCPServer, requiring mcp[cli]>=2.1.1,<3.0; the MCP-for-Stata package remains on the 1.x release line
• Protocol Compatibility: Support modern 2026-07-28 negotiation while retaining legacy 2025-11-25 initialize-era clients
• Data Info Performance: Reuse one parsed DataFrame within each handler instance so a single get_data_info request does not repeatedly read the same dataset; separate requests still use separate handlers
• Project Metadata: Clarify the security-first PyPI description and synchronize multilingual README feature/security content

Fixed

• Registry Metadata: Synchronize the MCP Registry server and PyPI package references with version 1.24.0

Security

• Security Linkage: Mark blocked stata_do, get_data_info, and read_log calls as blocked with executed: false and linked sanitized records in audit/security.jsonl
• Privacy: Redact credential-like fields and URL credentials/query/fragment data from persistent diagnostics; never persist dangerous do-file command text in the security ledger

Upgrade Notes

• Python 3.11+ is still required. The SDK upgrade does not rename the existing MCP tools or CLI entry point.
• Stata discovery is limited to macOS and Windows; existing Linux server and execution functionality remains available. Discovery checks installation files, not license validity, and may miss unregistered custom installations outside the searched locations.
• New do-file snapshots use .statamcp/snapshot/objects/<full-sha256>.do, with per-run metadata in .statamcp/snapshot/metadata.jsonl.
• Local debug tracing is enabled by default. Disable it with [DEBUG.tracing] ENABLED = false or STATA_MCP__DEBUG_TRACING_ON=false; audit records and snapshots are separate from rotating debug files.
• get_data_info remains beta-gated on Windows by default; this release does not remove that restriction.

Click to expand v1.23.1 details

[1.23.1] - 2026-08-29

Security

• Dependencies: Raise the minimum supported versions of MCP, cryptography, and pyasn1 to avoid known-vulnerable dependency ranges while retaining MCP 1.x compatibility
• GitHub Actions: Apply least-privilege permissions to lint and label-sync workflows and correct the label-sync trigger to use master

Added

• MCP Server: Add beta instructions for reliable Stata graph export and regression-table output
• OpenSSF: Display the OpenSSF Best Practices badge consistently across all README translations

Changed

• Release Workflow: Rename the PyPI publishing workflow display name without changing its release trigger

Click to expand v1.23.0 details

[1.23.0] - 2026-08-27

Security

• Windows Stata Do: Prevent crafted do-file names from reaching cmd.exe through temporary wrapper paths by generating UUID-based wrapper names independent of user-controlled filenames
• Windows Stata Do: Create temporary wrapper files atomically and verify cleanup after launch errors and monitored timeouts

Added

• WorkBuddy: Add installer and verifier support for WorkBuddy, including the wb alias and preservation of existing MCP server entries
• Pi: Add explicit-only Pi setup that installs pi-mcp-adapter when available, otherwise prepares ~/.pi/agent/mcp.json and reports that the integration is not yet active
• GitHub Copilot CLI: Add installer and verifier support through copilot mcp add, with an isolated ~/.copilot/mcp-config.json fallback

Changed

• Citations: Remove pinned package versions from citation metadata and README examples so citations remain valid across releases

Fixed

• Docs: Correct DeepSeek Harness documentation rendering on the project site

Click to expand v1.22.0 details

[1.22.0] - 2026-08-13

Added

• DeepSeek Harness: Add Day 0 support with equivalent dsh and deepseek-harness installer targets, a dedicated DSH patch entry, a 20-minute tool timeout, and duplicate protection based on id: stata-mcp
• Data Info Cache: Add a versioned JSON Schema document with strict JSON serialization and cache compatibility checks
• MCP Server: Add richer system instructions and an available-version hint

Changed

• Ado Install: Avoid ctx.elicit so clients without MCP elicitation support can use the explicitly enabled unsafe profile
• Dependencies: Allow uv-build versions from 0.11.28 up to, but not including, 0.13

Fixed

• Data Info Cache: Preserve the flat cache shape, normalize non-finite values, and accept numeric Excel column names

Click to expand v1.21.4 details

[1.21.4] - 2026-08-05

Fixed

• Installer: Keep installer's shell window open on errors, discover uv from standard paths, and handle piped execution safely

Click to expand v1.21.3 details

[1.21.3] - 2026-07-30

Fixed

• Dependencies: Pinned mcp[cli] to <2.0 to prevent unexpected breaking changes from a future major version release

Click to expand v1.21.2 details

[1.21.2] - 2026-07-28

Added

• Configuration: BETA.enable_windows_data_info option and STATA_MCP__ENABLE_WINDOWS_DATA_INFO environment variable to opt into the get_data_info MCP tool on Windows
• MCP Server: Windows-only beta gate that hides get_data_info from the MCP tool list unless the beta flag is enabled

Changed

• MCP Server: get_data_info is no longer advertised on Windows by default while the Windows-specific MCP wrapper bug is being investigated

Click to expand v1.21.1 details

[1.21.1] - 2026-07-16

Added

• Data Info: DATA_INFO_IS_CACHE config option to disable get_data_info caching
• Diagnostics: Privacy-safe stage timing, per-process debug logs, and watchdog stack snapshots for get_data_info investigations

Changed

• Diagnostics: Renamed Windows-specific diagnostic identifiers so the new logging is available on all platforms when IS_DEBUG is enabled

Click to expand v1.21.0 details

[1.21.0] - 2026-07-11

Added

• Async Stata Do: Beta async stata_do execution with concurrency limits for multi-client workflows
• Configuration: Layered config files (project, user, Linux system) with unified override resolution
• Security: Audit logging for security rejections, boundary violations, and lifecycle events
• Security: Static dofile parser that extracts data paths and expands macros for guard path auditing
• Security: strict_read_log_boundary switch and hardened get_data_info path/URL validation
• Build: Nuitka-based release binaries with wheel and sdist included in release assets
• Skill: New mcp-smoke-test skill for local MCP server smoke testing

Changed

• MCP Server: Removed deprecated write_dofile tool registration
• Dependencies: Dropped unused python-dotenv and pathvalidate from core dependencies
• CI: Dropped macOS x86_64 binary build and fixed Windows binary build

Fixed

• SPSS: Use DEFAULT_TIMEOUT for SPSS URL downloads
• Docs: Corrected stata_do examples and removed references to non-existent features

Click to expand v1.20.2 details

[1.20.2] - 2026-07-07

Security

• Stata Log: Capped SMCL log expansion at 10,000 characters for {hline} and {space} directives to prevent memory exhaustion from malicious log content

Changed

• CLI: Removed stale help parameters (--enable-smcl, --is-read-log) and simplified --replace help text
• Data Info: Added explicit > env > config > default resolution for string_keep_number, decimal_places, and hash_length; synced supported format list across tool description
• Project: Updated homepage and documentation URLs in pyproject.toml and __init__.py
• Dependencies: Bump pydantic-settings from 2.14.1 to 2.14.2

Documentation

• Added setup guides for Codex and Claude Science
• Added troubleshooting guide and migrated uv mirror documentation
• Updated README comparison tables, tagline, and project links across language versions
• Updated AGPL 3.0 license explanation and security reporting policy

Click to expand v1.20.1 details

[1.20.1] - 2026-06-19

Security

• Guard: Stripped Stata block comments (/* */) before security validation to prevent dangerous commands from being hidden inside comments
• Guard: Stripped frame prefixes before command checks so dangerous commands cannot bypass detection via frame scoping

Changed

• Dependencies: Updated locked dependencies to resolve security alerts

Click to expand v1.20.0 details

[1.20.0] - 2026-06-14

Fixed

• Data Info: Fixed crash when summarizing string columns containing numeric values across DTA, CSV, and XLSX formats

Added

• Tests: Added cross-format tests for string-numeric column handling and base type detection

Click to expand v1.19.0 details

[1.19.0] - 2026-06-09

Security

• Ado Install: Moved ado_package_install behind the unsafe MCP server profile; the tool is no longer available in core or all profiles, requiring explicit opt-in via --unsafe
• Ado Install: Added entry-point approval layer so stata-mcp install must explicitly approve ado installation capability before the tool becomes usable
• Ado Install: Enforced alphanumeric package name validation to prevent command injection through malformed package names
• Ado Install: Closed multiple security bypasses in the ado installation flow, including direct loader validation and centralized Stata command input validation
• Stata Help: Hardened Stata help lookup with direct loader validation and refreshed help cache behavior to prevent unauthorized command execution paths

Added

• Stata Do: Added optional timeout parameter for stata_do execution to prevent runaway Stata processes

Changed

• Stata Do: Use Stata return codes for install status detection instead of string parsing
• Dependencies: Bump starlette from 1.0.0 to 1.0.1

Click to expand v1.18.2 details

[1.18.2] - 2026-06-04

Added

• Installer: New src/stata_mcp/utils/installer/output.py module with ColorStream and colored_stdout() for ANSI-colored install output.
  • ColorStream buffers writes by line and colorizes any line starting with a known [TAG]\t prefix.
  • colored_stdout() swaps sys.stdout for a ColorStream(sys.stdout) inside a with block and restores it on exit.
  • Tag-to-color map: [ERROR] red, [DONE] green, [WARN] yellow, [BACKUP] cyan; unknown prefixes pass through plain.
  • Honors NO_COLOR env var and disables itself when the underlying stream is not a TTY.
• Tests: 13 new tests in tests/test_color_stream.py covering tag matching, buffering, flush, NO_COLOR, non-TTY, context-manager restoration, exception safety, and __getattr__ passthrough.

Changed

• CLI: handle_install now wraps each Installer.install*() call with colored_stdout() and prefixes its own print(...) lines with [ERROR] / [DONE] / [WARN] so the same wrapper handles both handler- and core-layer output.
• Verifier: _paint in verifier.py now delegates to the shared _should_color(stream) and _wrap(code, text) helpers in output.py, removing the duplicated NO_COLOR + isatty() check.
• Tests: tests/test_install.py updated to assert on [WARN]\t prefixes and adds an end-to-end test that verifies ANSI escapes for all four tags plus a plain ✅ line.

Click to expand v1.18.1 details

[1.18.1] - 2026-06-03

Changed

• Refactor: Rename src/stata_mcp/utils/Installer/ to lowercase src/stata_mcp/utils/installer/ for PEP 8 conformance.
• Refactor: Move src/stata_mcp/utils/Verifier/verifier.py into src/stata_mcp/utils/installer/verifier.py since verification is conceptually part of the install workflow.
• Refactor: Re-export Verifier, VerifyOutcome, VerifyResult, and color helpers from utils/installer/__init__.py.
• Refactor: Update import paths in cli/_handlers.py and three test files (test_install.py, test_installer_backup.py, test_verify.py).

Click to expand v1.18.0 details

[1.18.0] - 2026-06-03

Added

• Verify: New stata-mcp verify subcommand for read-only diagnostic checks of MCP client config files
  • verify -c <client> checks built-in clients (codex, claude, cc, claude-code, gemini, cursor, cline, opencode, openclaw, hermes, hermes-agent)
  • verify -f <path> [--index mcp.servers] [--key stata-mcp] checks custom JSON or TOML files
  • Exit codes: 0 (verified/warning), 1 (file/path/permission), 2 (key/index), 3 (parse), 4 (schema), 5 (usage)
  • Stdio warning flags localhost-like commands or http:// prefixes
  • Argparse usage errors exit 5 (overridden from default 2)
  • --key defaults to stata-mcp; accepts any custom key inside the target dict
• Installer: Add find_config_path(client) and find_default_index(client) read-only helpers to the Installer class
• Tests: 58 new tests in tests/test_verify.py covering success paths, parse errors, schema errors, warnings, permission-denied, and CLI routing

Changed

• Rebrand: Project renamed to MCP-for-Stata per StataCorp request
  • pyproject.toml: new description, expanded keywords, full trove classifiers
  • README*.md: hero, tagline, and quick-start sections updated across all four languages
  • CITATION.cff: title and repository updated to mcp-for-stata

Click to expand v1.17.3 details

[1.17.3] - 2026-05-20

Security

• Stata Do: Fix critical log_file_name injection vulnerability that allowed arbitrary Stata command execution via crafted log file names (closes #74).
• Stata Do: Add dofile_path control character validation to prevent quote and backtick injection in Stata do commands.
• Stata Do: Add log path traversal protection using resolve() and is_relative_to() checks.
• Guard: Harden validator against colon-prefix bypasses (quietly:, capture:, noisily:).
• Guard: Reject #delimit ; syntax in guard mode to prevent delimiter-based command smuggling.
• Guard: Enhance macro expansion detection to catch local/global macro bypasses with arguments and compound quotes.
• Guard: Add python, mata, java, and plugin commands to the dangerous commands blacklist.

Added

• Tests: Add comprehensive regression tests for all security fixes in test_guard_validator.py and test_stata_do_boundary.py.
• Docs: Add Codex installation instructions to README and Chinese translation.

Click to expand v1.17.2 details

[1.17.2] - 2026-05-19

Fixed

• Update: Fix uv-tool installation detection in stata-mcp update by using sys.prefix instead of sys.executable.resolve() to correctly identify uv tool environments (closes #69).

Added

• Tests: Fetch fixture data from Stata Press at runtime for improved test reliability.
• Tests: Add install parser and installer backup tests.

Changed

• Chore: Remove tests/ from .gitignore to include test files in version control.

Click to expand v1.17.1 details

[1.17.1] - 2026-05-18

Added

• Install: Backup original config file before editing during stata-mcp install
  • Backup naming: stem.backup-yyyymmddHHMM.suffix in the same directory
  • Covers JSON, TOML, and YAML config paths
  • Aborts install if backup fails to prevent data loss

Click to expand v1.17.0 details

[1.17.0] - 2026-05-16

Added

• Install: Custom JSON path and nested key support via --json-file and --json-index flags
  • --json-file PATH writes MCP config to an arbitrary JSON file
  • --json-index KEY.PATH specifies nested keys (e.g. mcp.servers for OpenClaw)
  • Bare stata-mcp install now defaults to --all instead of Claude Desktop only
• Install: CLI-first installation fallback for supported clients
  • Claude Code: tries claude mcp add before falling back to ~/.claude.json
  • Codex: tries codex mcp add before falling back to TOML config
  • OpenClaw: tries openclaw mcp set before falling back to ~/.openclaw/openclaw.json
  • Hermes Agent: tries hermes mcp add before falling back to ~/.hermes/config.yaml
• Install: New client support for Hermes Agent with YAML config generation
• Config: stata-mcp config show now supports Section.Key notation (e.g. DEBUG.IS_DEBUG)

Changed

• CLI: Flattened config subcommand structure (stata-mcp config show / set / cli set)
• Docs: Synchronized all multilingual READMEs, CLI docs, and agent guides with new install features

Click to expand v1.16.3 details

[1.16.3] - 2026-05-12

Added

• Install: Add OpenClaw client support via stata-mcp install -c openclaw, writing config to ~/.openclaw/openclaw.json.

Changed

• Dependencies: Upgrade mcp to 1.27.1, psutil to 7.2.2, requests to 2.34.0, pyreadstat to 1.3.4, python-dotenv to 1.2.2.
• Dependencies: Move openai packages to optional agents group and mcp-ui-server to ui group.

Click to expand v1.16.2 details

[1.16.2] - 2026-05-04

Security

• Stata Do: Add directory boundary validation so dofiles can only be executed from STATA_MCP_FOLDER.DO or WORKING_DIR (PR #70).
• Guard: Strengthen blacklist with dangerous command abbreviations (sh, xsh, winex, unixc, era, rmd) (PR #72).
• Guard: Add local macro expansion detection to catch bypasses like local cmd "shell" followed by `cmd' (PR #72).
• Guard: Log [SECURITY] warning when Guard is disabled at startup/execution time (PR #72).

Click to expand v1.16.1 details

[1.16.1] - 2026-04-22

Fixed

• MCP Server: Disable help resource registration to fix startup crash caused by URI parameter mismatch between {cmd} and function signature (cmd, replace).

Click to expand v1.16.0 details

[1.16.0] - 2026-04-20

Added

• Doctor: Add check_cleanup to scan and optionally remove old log/dofile/tmp files older than CLEAN_LOG_DAYS (PR #66 closes #63).
• Config: Add STATA_MCP__FOLDER_TAG config with env var override (default: .statamcp) and migration warning for legacy stata-mcp-folder (PR #67).

Fixed

• Stata Do: Insert clear before Unix do-file runs to prevent Stata session hangs (PR #64 closes #60).
• API: Rename is_read_log to read_log_when_error for clearer semantics (PR #65 closes #59).

Changed

• MCP Server: Trim verbose tool docstrings for stata_do, ado_package_install, and get_data_info to reduce token usage (issue #58).

Click to expand v1.15.1 details

[1.15.1] - 2026-04-09

Changed

• Data Info: Changed head parameter default from 5 to 0 in get_data_info to prevent context overflow on large datasets. Preview rows are now opt-in instead of opt-out.
• Docs: Updated tool docs (tools.md, tools.zh.md) to reflect new head default value.

Click to expand v1.15.0 details

[1.15.0] - 2026-04-07

Performance

• MCP Server: Refactor to tool registry pattern with lazy imports for faster MCP server startup (🚀 🐰 💨 Yes, I have seen you say it is slow 👀)
  • New _TOOL_REGISTRY dict maps tool names to metadata (description, func, profiles, unix_only, deprecated)
  • register_tools() function registers tools based on selected profile (core or all)
  • core profile registers: stata_do, get_data_info, help
  • all profile registers all tools including read_log, ado_package_install, write_dofile
  • Tools no longer registered at import time; register_tools() must be called explicitly
• MCP Server: Improved tool descriptions for better LLM tool selection accuracy
• Package Init: Lazy __getattr__ access to stata_mcp server and main entry point
  • Avoids eager import of heavy MCP server dependencies at package import time
  • Graceful PackageNotFoundError fallback for development environments

Added

• Data Info: head parameter in get_data_info for data preview rows
  • Positive values return first N rows, negative values return last |N| rows, 0 disables
  • head_warning added when requested rows exceed actual data size
• Data Info: Cache full variables/dimensions in summary, filter on demand
  • Summary now computes all variables regardless of vars_list; filtering applied in info property
  • Cache lookup simplified to hash-only comparison (no longer checks var_list subset)
• Read Log: lines parameter for content trimming
  • Positive values return first N items, negative values return last |N| items, 0 returns all
  • dict format trims by entry count instead of text lines for consistent behavior
• CLI: stata-mcp server subcommand with --core / --all profile flags
  • Supports -t/--transport option for stdio/sse/http
• CLI: stata-mcp doctor diagnostics subcommand
  • Comprehensive health checks for Stata CLI, Python version, OS, config, dependencies
  • Supports --json, --verbose, --check options for selective checks
  • Replaces deprecated --usable flag
• CLI: stata-mcp update subcommand for one-command package updates
  • Auto-detects install method (pip, uv-tool, uvx, homebrew, editable)
  • Supports --check for version comparison, --dry-run for preview, --method for override
• Tests: Added test_server_parser.py (4 cases) and test_server_registration.py (7 cases)

Changed

• MCP Server: Help function moved out of if config.IS_UNIX: conditional block with lazy-load guard
• MCP Server: Help resource registration moved into register_tools() for consistency
• CLI: --usable flag now deprecated with DeprecationWarning (will be removed in v1.16.0)
• CLI: __init__.py modules use lazy imports to avoid heavy dependency loading at startup
• Data Info: load_cached_summary() returns full summary (no longer applies _filter_var)

Documentation

• Docs: Update CLI docs (cli.md, cli.zh.md), CLAUDE.md for server profiles, doctor, update commands
• Docs: Update tool descriptions in _TOOL_REGISTRY for all MCP tools

Click to expand v1.14.3 details

[1.14.3] - 2026-04-06

Performance

• MCP Server: Lazy import heavy dependencies (StataDo, StataHelp, GuardValidator, RAMMonitor, etc.) inside tool functions to reduce startup time
• Config: Cache parsed TOML config with @cached_property and invalidate on write via save_config()

Changed

• Logging: Use config.IS_DEBUG to control file log level instead of hardcoded logging.DEBUG
• Logging: Remove dead IS_DEBUG variable assignment and redundant ENABLE_WRITE_DOFILE module-level constant
• Config: Invalidate cached config after save_config() to ensure fresh reads

Click to expand v1.14.2 details

[1.14.2] - 2026-04-04

Performance

• Startup: Delay STATA_CLI resolution and cache found path to config
• Startup: Lazy-load help_cls to avoid resolving Stata path at import time
• Startup: Lazy-import data_info to avoid loading pandas/numpy at startup
• Startup: Lazy-create stata-mcp-folder on tool invocation instead of at import

Fixed

• Guard: Strip Stata prefixes before security validation
• Help: Guard help save path when project temp directory is unset
• Help: Fix help cache lookup when project temp directory is unset
• Monitor: Ensure monitors are stopped in _execute_windows_with_monitors finally block
• Monitor: Remove redundant monitor.stop() calls and ensure process cleanup
• StataDo: Prevent subprocess resource leaks in execution methods

Docs

• Fix tool name reference from ssc_install to ado_package_install

Click to expand v1.14.1 details

[1.14.1] - 2026-03-24

Fixed

• Ado Install: Fixed SSC install success detection to be case-insensitive for better reliability
• Ado Install: Added explicit error summary extraction when installation fails for clearer feedback
• Ado Install: Improved error messages for all package sources (SSC, GitHub, net)

Click to expand v1.14.0 details

[1.14.0] - 2026-03-20

Added

• API Module: Created new API package with standalone functions for CLI and external use
  • Added stata_do, stata_help, read_log, get_data_info, ado_install, write_dofile API functions
  • Added _runtime.py for shared runtime utilities
• CLI: Added API-backed CLI tool commands (stata-mcp help, stata-mcp do, stata-mcp log, stata-mcp data-info, stata-mcp install)
  • CLI tools now behave like one-shot helpers with proper exit codes

Changed

• Architecture: Promoted core modules to top-level
  • Moved stata_mcp/core/stata to stata_mcp/stata
  • Moved stata_mcp/core/data_info to stata_mcp/data_info
• CLI: Refactored CLI code structure for better maintainability
  • Split _cli.py into _parsers.py (argument parsing) and _handlers.py (command handlers)
  • Reduced _cli.py from 458 lines to a thinner orchestrator
• Stata: Moved stata_log module to builtin_tools directory
• Ado Install: Increased default timeout to 300s for package installation

Click to expand v1.13.42 details

[1.13.42] - 2026-03-19

Changed

• MCP Tools: Made write_dofile tool optional via ENABLE_WRITE_DOFILE flag (default: False)
  • Added ENABLE_WRITE_DOFILE config property for conditional tool registration
  • Converted write_dofile from decorator to conditional registration pattern
  • Added deprecation warning in tool docstring
  • Tool only registers when ENABLE_WRITE_DOFILE=True

Click to expand v1.13.41 details

[1.13.41] - 2026-03-17

Added

• CLI: Added config inspect and STATA_CLI set commands
  • New stata-mcp config inspect command to view configuration
  • New stata-mcp config set stata_cli <path> command to set Stata CLI path
• Data Info: Added unified URL fetch with timeout support
• Registry: Added server.json for MCP registry

Changed

• Config: Use tomli-w for TOML serialization (replacing manual string formatting)
• Data Info: Unified URL and local file handling with bytes_io_data
• Data Info: Reorganized metrics for better code structure
• Security: Removed unsafe code execution calls in ado_install and macos finder
• MCP Tools: Removed append_dofile tool and deprecated write_dofile
• Windows: Use list args with shell=True for safer path handling

Click to expand v1.13.40 details

[1.13.40] - 2026-03-03

Changed

• MCP Tools: Removed load_figure tool from MCP server (unused and causes errors for non-vision models)
• CI: Added missing GitHub Actions permissions (attestations: write, actions: read) for provenance-based publish flows
• Documentation: Updated tools and overview docs to reflect load_figure removal

Click to expand v1.13.39 details

[1.13.39] - 2026-03-01

Added

• CI: Added MCP registry publishing step to PyPI workflow

Changed

• MCP Tools: Removed mk_dir tool from MCP server
• MCP Tools: Added error notes for non-vision model
• Documentation: Updated CLAUDE.md environment variables to match actual code

Click to expand v1.13.38 details

[1.13.38] - 2026-02-28

Added

• CLI: Enhanced usable command with uv check support
  • Added check_uv() to detect uv installation and prioritize it over Python version check
  • Added print_info() for non-failure info messages
• Documentation: Added MCP registry signature badge to README

Changed

• CLI: Refactored usable command with config integration
  • Uses config.IS_UNIX instead of platform detection
  • Simplified check_os() implementation
  • Fixed env var name: stata_cli -> STATA_CLI
• Style: Improved multi-line string formatting in mcp_servers.py

Click to expand v1.13.37 details

[1.13.37] - 2026-02-26

Added

• CLI: Enhanced install command with multi-client support
  • Added --all flag for one-click installation to all supported clients
  • Added --json-file option to read installation config from JSON file
  • Added Gemini client support via --client gemini
  • Added OpenCode client support via --client opencode
• Log System: New stata_log module for Stata log file parsing
  • StataLogTEXT: Parser for plain text .log files
  • StataLogSMCL: Parser for SMCL (Stata Markup and Control Language) .smcl files
  • Factory pattern with StataLog class for automatic format detection
  • Closes #37
• Data Info: Added SPSS file support (.sav, .zsav)
• Dependencies: Added pyreadstat>=1.2.0 and requests>=2.28.0 for SPSS support

Changed

• MCP Tools: Renamed read_file to read_log with StataLog integration
• Dependencies: Upgraded pandas to >=3.0.0,<4.0.0 (closes #36)
• Style: Unified function parameter formatting in mcp_servers.py

Fixed

• Data Info: Fixed HASH_LENGTH type error and URL hash bug
• CLI: Fixed --all flag using install_all() instead of install("all")

Documentation

• Added CI/CD status badges (PyPI publish, Docker build) to README
• Updated tool docs for read_log with StataLog integration
• Updated stata_do docstring for SMCL support

Click to expand v1.13.36 details

[1.13.36] - 2026-02-23

Fixed

• Guard: Fixed false positives in security validation
  • Added line-start anchor (^\s*) to all DANGEROUS_PATTERNS to prevent matching commands in the middle of lines
  • Fixed describe, summarize, and capture confirm variable rm being incorrectly flagged as dangerous
  • Commands like rm, do, shell are now only matched at line start (allowing indentation)
  • Removed duplicate \bdo\s+.* pattern

Click to expand v1.13.35 details

[1.13.35] - 2026-02-22

Added

• Data Info: Variable labels support for Stata .dta files
  • get_data_info now returns variable labels in vars_detail for DTA files
  • Added _get_var_extra_info hook method in base class for subclass extensibility

Fixed

• Data Info: Unified extension validation across CSV and Excel handlers
  • Both now use self.suffix and self.supported_extensions consistently
  • Removed redundant valid_extensions local variables

Changed

• CLI: Added --source option for Docker image registry selection
• Docker: Detect architecture instead of OS for platform flag

Documentation

• Enhanced mkdocs config with i18n and theme features
• Added Chinese translations for i18n support
• Added collapsible new features section and fixed usage links

CI

• Added mkdocs-static-i18n to gh-pages workflow

Click to expand v1.13.34 details

[1.13.34] - 2026-02-16

Changed

• Docker Refactor: Complete Docker image architecture redesign
  • Now using AEA Data Editor's Stata images as base
  • No longer distributes Stata installer - users only need to provide license file
  • New image naming convention: stata-mcp_{VERSION}_{EDITION} (e.g., stata-mcp_19_5_mp)
  • Supports StataNow 19.5, 18.5, and Stata 18 with mp/se/be editions
  • 9 image variants built via GitHub Actions matrix

Added

• CLI: Enhanced sandbox-install command with new options
  • --version / -V: Specify Stata version (19_5, 18_5, 18)
  • --edition / -e: Specify Stata edition (mp, se, be)
  • --tag: Specify Docker image tag (default: latest)
  • Example: stata-mcp sandbox-install -V 19_5 -e mp -l /path/to/stata.lic

Fixed

• Docker build permissions and lowercase repository name issues in GitHub Actions

Documentation

• Updated Docker guide with new image naming and CLI options
• Added license notice acknowledging StataCorp and AEA Data Editor
• Added recommendation note for Docker beginners

Click to expand v1.13.33 details

[1.13.33] - 2026-02-15

Added

• Docker Support: Full Docker containerization support for Stata-MCP
  • Added Dockerfile for containerized Stata-MCP deployment
  • Added GitHub Actions workflow for Docker image build and push to GHCR
  • Supports Stata 17/18/19 with configurable installer URL
• CLI: New sandbox-install command for Docker-based installation
  • Install Docker-based Stata-MCP to MCP clients (Claude, Cursor, Cline, etc.)
  • Configurable options: license file, work directory, CPU/memory limits, image tag
  • Usage: stata-mcp sandbox-install -l /path/to/stata.lic -c claude

Changed

• Dockerfile: Use local source code instead of PyPI for development builds

Documentation

• Docker Guide: Added comprehensive Docker documentation (docs/docker.md)
  • Image pull instructions for GHCR and DockerHub
  • Container running examples with parameter explanations
  • MCP client configuration for Docker-based setup
  • Building from source with local/remote installers
• Index: Added Docker to Advanced Features section

Click to expand v1.13.32 details

[1.13.32] - 2026-02-06

Changed

• CLI Refactor: Redesigned CLI structure with subcommands for better usability
  • Replaced stata-mcp --agent with stata-mcp agent run subcommand
  • Replaced stata-mcp --install with stata-mcp install subcommand
  • Added support for -c/--client option to specify installation target (claude, cc, cursor, cline, codex)
  • Set -t/--transport default to stdio explicitly
• Documentation: Comprehensive documentation updates
  • Added new CLI reference guide (docs/cli.md) with complete command documentation
  • Updated all existing documentation with new CLI syntax
  • Updated README and translations (cn, fr, sp) with new command examples
  • Updated mkdocs.yml navigation to include CLI Reference

Breaking Changes

• CLI Commands: Flag-based commands changed to subcommands
  • stata-mcp --agent → stata-mcp agent run
  • stata-mcp --install [client] → stata-mcp install -c [client]
  • Users will need to update scripts and documentation using old syntax

Click to expand v1.13.31 details

[1.13.31] - 2026-02-03

Fixed

• Code Quality Improvements: Fixed 8 code issues discovered through AI code reviews
  • ado_package_install: Made source parameter case-insensitive to match documentation (e.g., "GitHub" now works)
  • logging.disable(): Added missing required level argument to prevent TypeError when logging is disabled
  • Windows Path Handling: Removed unicode_escape processing that corrupted Windows paths (e.g., C:\new\data → C:\n + ew)
  • Config Directory: Unified configuration directory from inconsistent ~/.stata_mcp and ~/.stata_mcp to single ~/.statamcp
  • write_dofile: Added microsecond precision to timestamps (from %Y%m%d%H%M%S to %Y%m%d%H%M%S%f) to prevent filename collisions
  • Code Comments: Translated Chinese code comments to English for international team maintainability

Removed

• Deprecated Modules: Removed deprecated code to simplify codebase
  • Removed claude_proj module (entire directory, 261 lines)
  • Removed StataFinderOLD class (154 lines of dead code)

Technical

• Code Review: Issues were discovered through multiple AI code review tools (ChatGPT, Codex, and other reviewers)
• Breaking Change: Config directory changed from ~/.stata_mcp to ~/.statamcp - users may need to migrate configurations
• Performance: Microsecond precision in timestamps reduces collision probability from 1/second to 1/microsecond (1,000,000x improvement)

Click to expand v1.13.30 details

[1.13.30] - 2026-02-02

Fixed

• macOS StataNow BE Detection: Fixed StataNow BE not being detected on macOS
  • StataNow BE uses CamelCase executable filename (e.g., StataBE) instead of lowercase (e.g., stata-be)
  • FinderMacOS now tries multiple possible executable name formats:
    • stata-{edition} (traditional lowercase format)
    • Stata{EDITION} (StataNow CamelCase format)
    • Stata{Edition} (alternative capitalization)
  • Ensures compatibility with StataNow BE/SE/MP while maintaining backward compatibility
  • Related to issue #35

Technical

• Compatibility: Extended Stata finder to support both traditional and StataNow naming conventions
• Backward Compatibility: Existing Stata installations continue to work (first option tried is original format)

Click to expand v1.13.29 details

[1.13.29] - 2026-02-01

Fixed

• Environment Variable Documentation: Fixed inconsistent environment variable names in documentation

  • Corrected debug mode environment variable from STATA_MCP_DEBUG to STATA_MCP__IS_DEBUG in usage docs
  • Fixed working directory environment variable documentation showing duplicate names
  • Clarified that STATA_MCP__CWD (double underscore) is the new standard
  • Added deprecation notice for legacy STATA_MCP_CWD (single underscore) variable
  • Documentation now matches actual code implementation

• Stata CLI Error Detection: Added early error detection when Stata CLI cannot be found

  • Created new StataCLINotFoundError exception class for clear error messages
  • Modified Config.STATA_CLI to check if finder returns None and raise exception immediately
  • Added helpful configuration instructions in error message
  • Previously, None was propagated to execution stage causing confusing runtime errors
  • Now raises at MCP server startup with actionable guidance

• Logging Configuration: Fixed file logging configuration not being respected

  • Removed fallback logic that forced file logging when no handlers existed
  • Changed condition from len(logging_handlers) == 0 or LOGGING_FILE_HANDLER_ON to just LOGGING_FILE_HANDLER_ON
  • File logging is now only enabled when explicitly configured
  • Previously, LOGGING_FILE_HANDLER_ON=false did not prevent file log creation when console logging was disabled

• Security and Path Handling: Changed Unix-like systems to use shell=False for subprocess execution

  • Changed subprocess.Popen shell parameter from True to False in macOS/Linux execution paths
  • Fixed incorrect usage of shell=True with list arguments (only first element was used)
  • Paths with spaces now work correctly on Unix-like systems
  • Eliminated potential shell command injection risk
  • Updated comments to reflect correct subprocess usage pattern
  • Windows execution remains unchanged (correctly uses shell=True with string arguments and manual quoting)

Technical

• Security: Improved subprocess security by avoiding shell parsing on Unix-like systems
• Reliability: Early error detection prevents confusing runtime failures
• User Experience: Configuration now behaves as documented and expected
• Documentation: Environment variable documentation now matches implementation

Click to expand v1.13.28 details

[1.13.28] - 2026-01-26

Changed

• Data Handler Registry Pattern: Implemented registry pattern for extensible data format handler management

  • Added DATA_INFO_REGISTRY in base.py for automatic extension-to-handler mapping
  • Implemented __init_subclass__ for automatic subclass registration with their supported extensions
  • Added get_data_handler() function for retrieving handlers by file extension
  • Replaced hardcoded CLASS_MAPPING with dynamic registry in get_data_info()
  • Added supported_extensions class attribute to all data handler classes (CsvDataInfo, DtaDataInfo, ExcelDataInfo)
  • Improved imports organization in mcp_servers.py with relative imports
  • Enhanced extensibility - new data formats can be added without modifying core get_data_info function

• Code Naming Consistency: Renamed internal base module files for better naming conventions

  • Renamed _base.py to base.py in agent_as_rag module
  • Renamed _base.py to base.py in data_info module
  • Updated all import statements to reflect new file names
  • Removed leading underscore convention as these are not private modules

Removed

• Sandbox Module Cleanup: Removed unused sandbox infrastructure code
  • Cleaned up 136 lines of unused sandbox module code
  • Removed sandbox README and core module implementations
  • Removed jupyter_manager implementation
  • Streamlined codebase by removing experimental features

Technical

• Code Architecture: Enhanced extensibility through registry pattern implementation
• Maintainability: Improved code organization with consistent naming conventions
• Developer Experience: Easier to add new data format support through registration mechanism

Click to expand v1.13.27 details

[1.13.27] - 2026-01-25

Added

• Performance Monitoring System: Complete RAM monitoring framework for Stata subprocess execution

  • Abstract MonitorBase class for extensible monitor implementations
  • RAMMonitor implementation with psutil integration for cross-platform RAM tracking
  • Daemon thread-based monitoring with configurable check interval (0.5s default)
  • Automatic process termination when RAM exceeds configured limit
  • Custom exception hierarchy: StataMCPError → RAMLimitExceededError

• Configuration System: Comprehensive monitoring configuration via config and environment variables

  • IS_MONITOR global toggle for enabling/disabling monitoring (default: false)
  • MAX_RAM_MB configuration property for RAM limit in MB (default: -1, no limit)
  • Environment variable STATA_MCP__IS_MONITOR for global toggle
  • Environment variable STATA_MCP__RAM_LIMIT for RAM limit configuration
  • Configuration priority: environment variables > config file > defaults

• Monitor Integration: Clean separation between monitored and non-monitored execution paths

  • Original execution functions remain unchanged (_execute_unix_like, _execute_windows)
  • New monitored execution functions (_execute_unix_like_with_monitors, _execute_windows_with_monitors)
  • execute_dofile() routes to appropriate path based on IS_MONITOR flag
  • Monitor list passed to StataDo constructor for flexibility
  • 100% backward compatible when IS_MONITOR=false

Changed

• StataDo Architecture: Enhanced with monitor support and improved execution control

  • Added monitors parameter to StataDo.__init__() for optional monitor instances
  • Added IS_MONITOR boolean property for efficient execution path selection
  • Maintains original behavior when no monitors provided
  • Clean separation of concerns with dedicated monitor execution paths

• MCP Server Integration: Wired up RAM monitoring from configuration

  • RAMMonitor instantiation based on config.IS_MONITOR and config.MAX_RAM_MB
  • Monitor list passed to StataDo constructor during initialization
  • RAMLimitExceededError caught and returned as error response to MCP client
  • Nested configuration check for cleaner monitor activation logic

• Configuration Enhancement: Added monitor configuration section to config system

  • New [MONITOR] section in config.example.toml
  • IS_MONITOR and MAX_RAM_MB configuration options documented
  • Default configuration ensures monitoring is disabled unless explicitly enabled

Technical

• New Dependencies: Added psutil>=6.0.0 for cross-platform process monitoring
• Code Quality: Clean architecture with abstract base class pattern for extensibility
• Backward Compatibility: Original functions unchanged, monitoring opt-in via configuration
• Performance: Minimal overhead when monitoring disabled, daemon thread when enabled
• Error Handling: Custom exception with detailed RAM usage information
• Documentation: Comprehensive docstrings noting feature is "coded by Claude Code with GLM-4.7"

Notes

• This is a beta feature with monitoring disabled by default
• When IS_MONITOR=false, behavior is 100% identical to previous versions
• When IS_MONITOR=true and MAX_RAM_MB is set, Stata processes are monitored and killed if RAM exceeds limit
• -1 in config for MAX_RAM_MB means no limit (converted to None)
• Monitor system designed for extensibility to support timeout monitoring in future

Click to expand v1.13.26 details

[1.13.26] - 2026-01-25

Fixed

• Windows Log File Completeness: Added capture log close before opening log to prevent incomplete log files
  • Ensures proper log file closure in both Unix and Windows environments
  • Adds working directory (cwd) parameter to StataDo for better execution control
  • Sets cwd in subprocess calls for proper execution context

Changed

• Windows Temporary File Management: Optimized temporary batch dofile handling on Windows

  • Changed temp directory from user directory to system temp directory
  • Uses dofile stem as temporary batch filename instead of timestamp
  • Removed dofile_base_path parameter from StataDo class (no longer needed)

• Code Quality Improvements: Enhanced StataDo implementation with better type safety

  • Refactored do.py to use Path types instead of strings
  • Replaced repetitive OS checks with boolean is_unix flag
  • Improved error handling with exception handling in read_log method

• Security Module Integration: Replaced validation logic with guard module

  • Replaced _validate_dofile_content with guard module integration
  • Centralized security validation for better maintainability

• Configuration Enhancements: Improved configuration and logging system

  • Re-constructed path configuration and IS_GUARD handling
  • Added back logging for output_base_path for better debugging
  • Updated config.example.toml with new configuration options

Technical

• Code Quality: Improved type safety with Path types and boolean flags
• Error Handling: Enhanced exception handling throughout StataDo
• Maintainability: Centralized validation logic in guard module
• User Experience: Better log file completeness and temporary file management

Click to expand v1.13.25 details

[1.13.25] - 2026-01-22

Added

• Configuration File Support: Added support for reading configuration from ~/.statamcp/config.toml file
  • Config priority: Environment variables (highest) > TOML config file > Default values (lowest)
  • Added Config class in src/stata_mcp/config.py with unified configuration management
  • Type-safe converters: _to_bool(), _to_int(), _to_path() with validators
  • Configuration properties: IS_DEBUG, LOGGING_ON, LOGGING_CONSOLE_HANDLER_ON, LOGGING_FILE_HANDLER_ON, LOG_FILE, MAX_BYTES, BACKUP_COUNT, SYSTEM_OS, IS_UNIX, STATA_CLI
  • Added config.example.toml for user configuration reference

Changed

• mcp_servers.py: Simplified configuration initialization using centralized Config class
  • Removed scattered environment variable reading logic
  • Delegated platform detection and Stata CLI finding to Config class

Click to expand v1.13.24 details

[1.13.24] - 2026-01-21

Added

• Security Guard System: Implemented comprehensive security validation for Stata dofile execution
  • Added guard package with blacklist-based security validation
  • blacklist.py: Defines dangerous commands (shell escapes, file operations, external code execution)
  • validator.py: Core validation logic with GuardValidator, RiskItem, and SecurityReport classes
  • Integrated guard into stata_do function to validate dofiles before execution
  • Returns detailed warnings and blocks execution when dangerous commands are detected
  • Configurable via STATA_MCP__IS_GUARD environment variable (default: enabled)

Security

• Blacklist Coverage: Protects against dangerous commands including:
  • Shell escapes: !, !!, shell, xshell, winexec, unixcmd
  • File operations: erase, rm, rmdir, copy
  • External code execution: run, do, include

Testing

• Added comprehensive test suite with 15 test cases covering all security scenarios
• Tests validate dangerous command detection, line number tracking, and edge cases

Click to expand v1.13.23 details

[1.13.23] - 2026-01-13

Added

• Multi-Agent Client Support: Extended installer with support for multiple AI agent platforms
  • Added support for Cursor IDE
  • Added support for Cline (VS Code extension)
  • Added support for Codex with TOML-based configuration

Changed

• Installer Refactor: Enhanced installer architecture for better extensibility
  • Moved command, args, env to instance variables for reusability
  • Added install_to_toml_config method for TOML-based configurations
  • Simplified configuration management across different clients

Technical

• TOML Support: Added tomllib import and TOML formatting helpers

Click to expand v1.13.22 details

[1.13.22] - 2026-01-12

Added

• Multi-Client Installation Support: Enhanced installer to support multiple client targets
  • Added install_to_json_config method for generic JSON config installation
  • Added install_to_claude_code method for Claude Code client support
  • Updated install method to accept target client parameter (claude/cc)
  • Support for installing to both Claude Desktop and Claude Code from CLI

Changed

• Environment Variable Configuration: Improved STATA_CLI configuration flexibility
  • Changed STATA_CLI and STATA_MCP_COMMON_CONFIG to @property methods
  • Support reading STATA_CLI from environment variable
  • Dynamic configuration loading for better portability

Enhanced

• CLI Installation: Enhanced command-line interface for installation
  • Simplified installation process with direct client target specification
  • Better user experience for installing to different clients

Click to expand v1.13.21 details

[1.13.21] - 2026-01-07

Security

• Path Traversal Protection: Enhanced file reading security for read_file tool
  • Restricted read_file to only access files within the stata-mcp-folder directory
  • Added path resolution and validation to prevent symlink-based attacks
  • Implemented security violation logging for audit trails
  • Raises PermissionError when attempting to access files outside allowed directory

Technical

• Security Hardening: Prevented unauthorized file system access through path traversal attacks
• Error Handling: Enhanced error messages with clear directory restrictions

Click to expand v1.13.20 details

[1.13.20] - 2026-01-06

Changed

• get_data_info Major Refactor: Comprehensive improvement to data information retrieval and processing
  • Added decimal_places parameter with environment variable support (STATA_MCP_DATA_INFO_DECIMAL_PLACES)
  • Enhanced summary statistics consistency across all data types
  • Added cached loading to summary method for improved performance
  • Simplified get_data_info function architecture with enriched output structure
  • Added string series method for better string variable handling

Fixed

• Metrics Filtering Bug: Correctly filter metrics in DataInfoBase._filter() method
  • Fixed metric filtering logic that was incorrectly processing data metrics
  • Ensures accurate metric selection and reporting in data info output

Added

• TSV and PSV Format Support: Extended data file format compatibility
  • Added support for Tab-Separated Values (TSV) files
  • Added support for Pipe-Separated Values (PSV) files
  • Updated mcp_servers to handle new delimiter-separated formats
  • Automatic delimiter detection for seamless file processing

Changed

• MCP Architecture Cleanup: Improved MCP server organization and maintainability
  • Reorganized tool functions order for better code structure
  • Removed all prompt-related code to streamline MCP server implementation
  • Cleaner separation of concerns in MCP server architecture

Technical

• Environment Variables: Added support for decimal precision configuration via environment variables
• Code Quality: Enhanced type annotations and function parameter design
• Performance: Improved summary method performance with cache integration
• Maintainability: Simplified codebase through removal of deprecated prompt features

Click to expand v1.13.19 details

[1.13.19] - 2026-01-04

Changed

• Cached Summary Loading: Enhanced summary method with automatic cache loading capability
  • Added load_cached_summary() integration in DataInfoBase.summary() method
  • Improved performance by loading from cache when available instead of reprocessing data
  • Maintains backward compatibility with existing functionality
  • Optimized data info retrieval workflow with cache-first approach

Technical

• Performance Enhancement: Reduced data processing time for repeated operations on same datasets
• Cache Architecture: Better integration of cached summary loading into main summary workflow

Click to expand v1.13.18 details

[1.13.18] - 2026-01-03

Changed

• Data Info Refactor: Simplified get_data_info function and improved architecture

  • Moved cache logic from function level to class level for better maintainability
  • Removed redundant cache lookup in get_data_info tool
  • Enhanced error handling and logging throughout data_info module
  • Improved code organization with clearer separation of concerns

• Series Classes Enhancement: Converted to dataclass for better structure

  • Refactored StringSeries and NumericSeries to use @dataclass decorator
  • Added configurable decimal_places parameter for numeric precision control
  • Added string_keep_number parameter to control value list display
  • Improved property-based implementation with type annotations

• Enriched Data Output: Enhanced get_data_info return structure

  • Added source field to overview section showing data file path
  • Added var_list to overview showing all variable names
  • Added info_config section with metrics, max_display, and decimal_places settings
  • Enhanced vars_detail with var field for each variable
  • Improved saved_path reporting for cache files

Added

• Environment Variable Configuration: Added support for data_info customization

  • STATA_MCP_DATA_INFO_DECIMAL_PLACES: Control decimal precision (default: 3)
  • STATA_MCP_DATA_INFO_STRING_KEEP_NUMBER: Control string value display (default: 10)

• Documentation: Updated get_data_info docstring with complete examples

  • Added detailed return value structure documentation
  • Added comprehensive examples showing string and numeric variable output
  • Improved clarity with full statistical output display

Technical

• Configuration Management: Centralized settings with environment variable support
• Code Quality: Improved type safety and code organization
• Documentation: Enhanced inline documentation and examples

Click to expand v1.13.17 details

[1.13.17] - 2026-01-01

Added

• Excel File Support: Complete support for Excel file formats (.xlsx, .xls) in get_data_info tool

  • New XlsxDataInfo class for handling Excel workbook metadata and statistics
  • Automatic detection and processing of Excel files with openpyxl integration
  • Support for both .xlsx and .xls file formats with comprehensive error handling
  • Enhanced data type detection for Excel-derived datasets
  • Improved file format validation for Excel workbooks

• Enhanced Format Support: Extended data_info tool with additional delimiter-separated formats

  • Added TSV (Tab-Separated Values) file format support
  • Added PSV (Pipe-Separated Values) file format support
  • Unified CSV/TSV/PSV handling in CsvDataInfo with automatic delimiter detection
  • Extended file extension validation for new formats
  • Improved data processing pipeline for varied text-based formats

Changed

• MCP Servers Update: Enhanced mcp_servers.py with comprehensive format support

  • Updated get_data_info tool to support Excel, TSV, and PSV formats
  • Improved file type detection and routing logic
  • Enhanced error messages for unsupported file formats
  • Better integration of new data format handlers

• Data Info Architecture: Improved extensibility for additional file formats

  • Enhanced base data_info classes for better format handling
  • Streamlined file extension validation across all formats
  • Improved type detection for Excel and delimiter-separated files

Technical

• New Dependencies: Added openpyxl>=3.1.5 for Excel file processing
• Code Quality: Enhanced file format detection with cleaner separation of concerns
• Type Safety: Improved type annotations for Excel and delimiter-separated file handling

Click to expand v1.13.16 details

[1.13.16] - 2025-12-31

Added

• Help Command Cache: Implemented local file caching for Stata help command responses
  • Automatically checks for cached help files before calling Stata
  • Cache files saved with format help__{command}.txt in temporary directory
  • Significantly reduces response time for repeated help requests
  • Improves efficiency by avoiding redundant Stata executions

Changed

• Centralized Cache Directory: Enhanced help system with centralized cache management

  • Added STATA_MCP_DIRECTORY constant for ~/.statamcp centralized directory
  • Help cache now stored in ~/.statamcp/help/ for better organization
  • Improved cache directory structure with dedicated help subdirectory
  • Better file management with centralized cache location

• Help System Refactoring: Moved help implementation to dedicated module

  • Extracted help logic from mcp_servers.py to stata_help.py
  • Added Help class for better code organization and maintainability
  • Improved separation of concerns with modular help implementation
  • Enhanced code structure for easier maintenance and testing

• Documentation Language: Updated help function documentation from Chinese to English

  • Translated Notes section in help function docstring to English
  • Improved international accessibility for English-speaking users
  • Consistent English documentation across the codebase

• Platform Detection: Enhanced IS_UNIX detection with explicit platform check

  • Replaced SYSTEM_OS.lower() != "windows" with explicit platform check
  • More reliable platform detection across different operating systems
  • Improved code clarity and maintainability

• Code Quality: Improved code formatting and style consistency

  • Applied linting rules to ensure consistent code style
  • Better code readability with proper formatting

Performance

• Dramatically improved response times for repeated help command requests
• Reduced redundant Stata help command executions
• Optimized help system workflow with cache-first approach

Click to expand v1.13.15 details

[1.13.15] - 2025-12-30

Added

• Enhanced Summary Statistics: Extended data info summary with three additional statistical measures
  • Added sample size (n) to indicate valid observation count for each variable
  • Added skewness (3rd moment) to measure distribution asymmetry
  • Added kurtosis (4th moment/excess kurtosis) to measure distribution tailedness
  • Provides more comprehensive statistical description for numerical variables
  • Updated docstring examples with real auto.dta dataset outputs

Changed

• Updated get_data_info tool to return 7 statistics instead of 4
• Enhanced docstring examples in both _base.py and mcp_servers.py
• Improved statistical analysis capabilities for data exploration

Click to expand v1.13.14 details

[1.13.14] - 2025-12-27

Added

• Help Command Cache: Implemented local file caching for Stata help command responses

  • Automatically checks for cached help files before calling Stata
  • Cache files saved with format help__{command}.txt in temporary directory
  • Significantly reduces response time for repeated help requests
  • Improves efficiency by avoiding redundant Stata executions

• Data Info Content-Based Caching: Implemented smart caching mechanism for data information

  • Uses MD5 hash of file content to generate unique cache identifiers
  • Cache filename format: data_info__{name}_{ext}__hash_{hash}.json
  • Ensures same filename and content produce identical cache files
  • Checks cache before processing, returns cached result if available
  • Supports HASH_LENGTH environment variable to customize hash length (default: 12)
  • Proper exception handling for corrupted or missing cache files

Performance

• Dramatically improved response times for repeated operations on same data files
• Reduced redundant Stata help command executions
• Optimized data info retrieval workflow with cache-first approach

Click to expand v1.13.13 details

[1.13.13] - 2025-12-25

Added

• StataNow Support: Enhanced macOS StataFinder to support StataNow directory detection

  • Added StataNow to the feature name search list with priority over Stata
  • Improved multi-version Stata installation detection for better compatibility
  • Returns list of found Stata installations and selects highest version via max()

• User Applications Directory: Extended Stata search paths to include user directory

  • Added ~/Applications to macOS search paths for user-installed Stata
  • Maintains priority: system /Applications > user ~/Applications
  • Better support for non-admin Stata installations

Changed

• Logging System Optimization: Enhanced logging infrastructure with improved debug control

  • Added IS_DEBUG flag for better debug mode control
  • Moved log file to ~/.statamcp/stata_mcp_debug.log for centralized management
  • Improved path handling with expanduser() and absolute() for reliability
  • Added parent directory creation with proper error handling
  • Added project name logging when using custom working directory

• MCP Server Initialization: Reorganized initialization order for better reliability

  • Moved MCP Server initialization after system checks
  • Removed Claude Code-specific client handling for cross-platform compatibility
  • Improved error message format with system information

Click to expand v1.13.12 details

[1.13.12] - 2025-12-23

Fixed

• Icon Fetch Timeout Issue: Added error handling for FastMCP icon initialization to prevent server startup failures
  • Implemented try-except block to gracefully fallback when icon URL fetch times out
  • Ensures server can start successfully even with network connectivity issues
  • Maintains full functionality regardless of icon loading status

Improved

• MCP Instructions Clarity: Enhanced server instructions for better LLM understanding
  • Simplified and optimized instructions text for clearer context
  • Emphasized do-file as the minimum operation unit
  • Clarified session management behavior (no session config)

Click to expand v1.13.11 details

[1.13.11] - 2025-12-17

Fixed

• Critical macOS Compatibility Issue: Resolved MCP server startup failures for macOS users

  • Fixed resource decorator parameter mismatch by updating URI from help://stata/{cmd} to help://stata/{cmd}?is_save={is_save}
  • Resolved ValueError that prevented server initialization due to function signature mismatch
  • Restored full functionality for macOS users upgrading from v1.13.9 to v1.13.10

• ADO Package Installation Logic: Improved platform-specific package management reliability

  • Replaced problematic try-except block with clean if-else platform separation
  • Removed complex exception handling that could cause silent failures or inconsistent error messages
  • Ensured stable behavior across Unix (macOS/Linux) and Windows platforms

Improved

• Cross-Platform Path Handling: Enhanced file path consistency across operating systems

  • Replaced str(path) with path.as_posix() in functions returning file paths
  • Updated results_doc_path, write_dofile, and append_dofile functions for POSIX-style path consistency
  • Improved reliability of file operations across different platforms

• Code Organization: Better structural organization for maintainability

  • Moved help class initialization inside IS_UNIX condition block for proper scoping
  • Removed unnecessary strict_mode parameter from write_dofile function
  • Enhanced code clarity and reduced potential configuration conflicts

Notes

• This release focuses on fixing critical compatibility issues introduced in v1.13.10
• All macOS users are strongly recommended to upgrade from v1.13.9 or earlier to this version
• Windows users continue to benefit from the package installation improvements introduced in v1.13.10

Click to expand v1.13.10 details

[1.13.10] - 2025-12-17

Added

• Help Function Enhancement: Added is_save parameter to help function for flexible help content management

  • New is_save: bool = True parameter allows users to control whether help content is saved to files
  • Help content is now automatically saved to tmp_base_path / f"help__{cmd}.txt" when is_save=True
  • Enhanced user experience with optional help content persistence for reference

• Windows Platform Support for ADO Package Installation: Enabled cross-platform Stata package management

  • Moved ado_package_install function from Unix-only condition block to all platforms
  • Windows users can now install Stata packages from SSC, GitHub, and net sources
  • Added Windows-specific error handling with fallback do-file execution for improved reliability
  • Updated __all__ exports to include ado_package_install for all platform users
  • Enhanced return type annotation for better IDE support and code clarity

Changed

• Code Simplification: Streamlined stata_do function log handling logic
  • Simplified conditional log content processing with ternary operator
  • Improved code readability without affecting business logic
  • Maintained identical functionality while reducing code complexity

Technical

• Cross-Platform Compatibility: Enhanced Windows platform support while maintaining Unix/Linux functionality
• Code Quality: Improved function parameter design with better type annotations and documentation
• User Experience: Expanded feature availability to all supported platforms regardless of operating system

Click to expand v1.13.9 details

[1.13.9] - 2025-12-13

Changed

• MCP Server Initialization: Simplified MCP server initialization and dependency management
  • Adjusted MCP version requirement from >=1.20.0 to >=1.16.0 for better compatibility
  • Removed version compatibility checks and complex error handling code
  • Cleaned up commented dependencies in pyproject.toml for cleaner configuration
  • Simplified FastMCP initialization with fixed icon configuration

Technical

• Dependency Management: Streamlined MCP version requirements and compatibility handling
• Code Simplification: Removed 53 lines of complex version checking and error handling code
• Initialization Performance: Enhanced server startup performance through simplified initialization logic

Click to expand v1.13.8 details

[1.13.8] - 2025-12-06

Changed

• Agent Architecture Modernization: Complete refactoring of agent mode with modern framework adoption

  • Implemented new REPLAgent using agents.as framework replacing deprecated LangChain-based implementation
  • Enhanced agent performance and reliability through modern agents.as integration
  • Updated CLI agent mode to leverage new REPLAgent for improved user experience
  • Removed 396 lines of deprecated code from langchain-based mode module

• Code Quality Enhancement: Significant codebase cleanup and modernization

  • Removed deprecated langchain-based mode module including agent, prompt generator components
  • Eliminated legacy PMP (Prompt Management Package) system for streamlined architecture
  • Improved maintainability through removal of obsolete code patterns and dependencies
  • Enhanced code organization with modern agent framework adoption

Technical

• Framework Migration: Transitioned from LangChain to agents.as framework for improved agent capabilities
• Performance: Enhanced agent execution efficiency and resource utilization
• Maintainability: Simplified codebase architecture with modern framework patterns
• Reliability: Improved agent mode stability through contemporary framework adoption

Click to expand v1.13.7 details

[1.13.7] - 2025-12-05

Changed

• System Detection Optimization: Enhanced system detection logic with IS_UNIX variable

  • Added IS_UNIX variable to replace repetitive SYSTEM_OS.lower() != "windows" checks
  • Optimized 4 conditional statements across the codebase for better performance
  • Improved code readability with simplified conditional logic
  • Enhanced maintainability with centralized system detection

• Module Interface Enhancement: Added explicit all list for better module exports

  • Introduced comprehensive all list defining public module interface
  • Categorized exports: Core Functions (get_data_info, stata_do, write_dofile, append_dofile)
  • Categorized exports: Utilities (mk_dir, load_figure, read_file)
  • Conditional exports for Unix-specific functions (help, ado_package_install)
  • Improved IDE support and module discoverability

Technical

• Performance: Reduced repetitive string operations with centralized system detection
• Code Quality: Enhanced module organization with explicit public interface definition
• Maintainability: Centralized system detection logic easier to modify and maintain
• Developer Experience: Better IDE autocomplete and module navigation with explicit exports

Click to expand v1.13.6 details

[1.13.6] - 2025-12-03

Fixed

• Windows Compatibility Issues: Critical fixes for Windows user experience and cross-platform support
  • Fixed environment variable name from STATA-MCP-CLIENT to STATA_MCP_CLIENT for proper Windows environment access
  • Disabled help functionality on Windows systems to prevent system crashes during help command execution
  • Disabled ado package installation on Windows systems to ensure stable operation
  • Enhanced cross-platform reliability with Windows-specific conditional code paths

Changed

• Function Parameters: Added strict_mode parameter to write_dofile function for enhanced control
• Code Quality: Fixed grammar typos and improved documentation clarity throughout code comments
• Platform Detection: Improved operating system detection with better case-insensitive handling

Technical

• Cross-Platform Architecture: Enhanced Windows support while maintaining functionality for macOS and Linux
• Error Prevention: Proactive disabling of problematic features on Windows to prevent crashes
• Documentation: Improved code comments and parameter documentation for better maintainability

Click to expand v1.13.5 details

[1.13.5] - 2025-11-26

Changed

• Logging System: Enhanced logging infrastructure with rotating file handler

  • Replaced FileHandler with RotatingFileHandler for better log management
  • Added 10MB file size limit with 5 backup files for automatic log rotation
  • Improved log file handling and disk space management

• Language Detection Optimization: Improved prompt language detection and configuration

  • Moved language detection logic from mcp_servers.py to utils/Prompt/init.py
  • Added STATA_MCP_PROMPT_LANGUAGE environment variable for manual language override
  • Enhanced system language detection with better fallback mechanisms
  • Improved code organization with centralized language management

Technical

• Code Quality: Enhanced import organization and removed unused locale import
• Configuration: Improved environment variable handling for better flexibility
• Reliability: More robust logging system with automatic rotation and backup management
• Maintainability: Better separation of concerns between logging and language detection

Click to expand v1.13.4 details

[1.13.4] - 2025-11-25

Fixed

• Log File Path Resolution: Fixed FileNotFoundError when running stata-mcp from directories without "~" subdirectory
• Path Configuration: Use Path.home() for log file path instead of hardcoded "~/.statamcp.log"
• User Experience: Eliminates need to manually create "~" folder in current directory

Click to expand v1.13.3 details

[1.13.3] - 2025-11-24

Added

• JSON Export Feature: Enhanced summary() method with saved_path parameter for exporting data summaries to JSON files
• UTF-8 Encoding Support: JSON files are saved with proper UTF-8 encoding for international character support
• File Path Tracking: Summary results now include the saved file path when export is used
• New save_to_json Static Method: Dedicated method for reliable JSON file writing with error handling
• Comprehensive Logging: Added detailed logging throughout the get_data_info function for better debugging and monitoring
• Enhanced Documentation: Updated function docstrings with complete output examples and usage patterns

Changed

• Core Function Refactoring: Complete restructure of get_data_info function implementation for better maintainability and performance
• Improved Data Processing: Optimized data processing pipeline with better separation of concerns
• Numerical Precision: Standardized to 3-decimal precision in summary outputs for consistency
• Error Handling: Enhanced error handling for file operations with proper boolean return status
• Documentation Examples: Added real data format examples in function documentation

Technical Improvements

• Code Organization: Better separation of concerns with dedicated helper methods
• Performance Enhancement: Optimized data processing pipeline for improved speed
• Path Validation: Proper handling of both string and PathLike objects for file paths
• Maintainability: Cleaner code structure for easier future development

Backward Compatibility

• ✅ All changes maintain full backward compatibility
• ✅ Existing code will continue to work without modifications
• ✅ New features are opt-in through additional parameters

Click to expand v1.13.2 details

[1.13.2] - 2025-11-23

Added

• Logging Control: Added controller for logging system
  • Environment variable control for flexible logging configuration
  • STATA_MCP_LOGGING_ON for complete logging enable/disable control
  • Separate handlers for console and file output with independent control
  • Improved user experience with customizable logging behavior

Click to expand v1.13.1 details

[1.13.1] - 2025-11-22

Added

• Comprehensive Logging System: Implemented complete logging infrastructure throughout MCP server
  • Added comprehensive logging system to MCP server for enhanced debugging and monitoring
  • Added logging to all 8 core tool functions for better operation tracking
  • Improved StataDo constructor documentation for clearer usage guidance

Changed

• Enhanced System Observability: Improved debugging capabilities with comprehensive logging coverage
  • Better visibility into MCP server operations and tool function executions
  • Enhanced error tracking and performance monitoring capabilities
  • Improved developer experience with detailed operation logs

Technical

• Logging Infrastructure: Complete logging system integration with configurable log levels
• Documentation: Enhanced constructor documentation for better developer understanding
• Debugging: Improved troubleshooting capabilities with comprehensive operation logging

Click to expand v1.13.0 details

[1.13.0] - 2025-11-20

Added

• Agent-as-Tools Framework: Complete framework for converting agents to MCP tools with seamless integration

  • New comprehensive agent conversion system for enhanced workflow automation
  • Flexible error handling and type safety throughout the agent-as-tools ecosystem
  • Enhanced module exports for improved discoverability and integration

• ADO Package Installation Framework: Comprehensive Stata package management system

  • Complete framework for installing Stata packages from multiple sources
  • SSC (Statistical Software Components) archive integration with enhanced search capabilities
  • Direct GitHub repository installation support for custom packages
  • Net install functionality for additional package distribution methods
  • Extensible architecture supporting future package sources and installation methods

• Multi-Provider API System: Enhanced model provider flexibility and management

  • New Provider class for flexible API provider management in set_model functionality
  • Support for multiple LLM providers with unified configuration interface
  • Improved API switching and configuration management for different provider environments

Security

• Shell-Escape Protection: Critical security enhancement to prevent OS command execution
  • Added validation step to block Stata shell-escape directives (!cmd, shell cmd)
  • Comprehensive input validation before do-file execution to prevent malicious commands
  • Enhanced security posture for production environments (Contributed by @123mutouren321414)
  • Fixes security issue #20 with robust protection mechanisms

Changed

• Pathlib Migration: Complete migration from os.path to pathlib.Path for enhanced cross-platform support

  • System-wide replacement of os.path operations with modern pathlib.Path interface
  • Improved cross-platform compatibility and path handling reliability
  • Enhanced type safety and better IDE support with path operations
  • More robust and maintainable path manipulation throughout the codebase

• Enhanced Error Handling: Comprehensive type safety improvements and parameter design enhancements

  • Fixed get_data_info tool write type errors for improved data processing reliability
  • Resolved get_exp_cwd to properly return Path object instead of string
  • Fixed dict -> str function return type errors throughout the application
  • Updated default args in get_data_info tool for better parameter design
  • Improved type annotations and validation across core functionality

• Code Architecture Enhancements: Improved maintainability and performance optimizations

  • Refactored help functionality into dedicated StataHelp class for better modularity
  • Enhanced log width issue resolution through terminal size environment configuration
  • Improved overall code organization with better separation of concerns
  • Enhanced performance and reliability through systematic code improvements

Fixed

• Type Safety Issues: Resolved multiple type-related errors for improved application stability

  • Fixed function return type errors that could cause runtime issues
  • Enhanced parameter validation and type checking across core tools
  • Improved error handling for edge cases and invalid input scenarios
  • Better integration between different components with consistent type handling

• Terminal and Logging Issues: Resolved display and logging problems for better user experience

  • Fixed log width issues by properly setting terminal size environment variables
  • Improved overall terminal output formatting and readability
  • Enhanced debugging capabilities with better error message formatting

Technical

• Code Quality: Comprehensive improvements to code maintainability and extensibility

  • Enhanced type annotations throughout the application for better IDE support
  • Improved error handling patterns with comprehensive exception management
  • Better separation of concerns with modular architecture enhancements
  • Enhanced documentation and code examples for improved developer experience

• Performance: Optimized application performance through systematic improvements

  • Faster startup times with optimized module loading and initialization
  • Improved memory usage patterns through better resource management
  • Enhanced overall application responsiveness and stability
  • Better integration efficiency between different system components

• Security: Strengthened security posture with comprehensive protection mechanisms

  • Enhanced input validation and sanitization throughout the application
  • Improved protection against common web and CLI security vulnerabilities
  • Better error handling that prevents information leakage in production environments
  • Comprehensive security review and hardening of core functionality

Click to expand v1.12.1 details

[1.12.1] - 2025-11-02

Added

• Linux Platform Support: Complete Linux platform Stata detection with intelligent directory search

  • New FinderLinux implementation with configurable search paths for Linux systems
  • Enhanced directory search in /usr/local/bin with automatic detection of Stata-related directories
  • Support for custom Stata installation locations through flexible path configuration
  • Full integration with existing StataFinder architecture maintaining cross-platform consistency

• Enhanced StataEditionConfig: Advanced comparison logic with comprehensive priority system

  • Priority-based sorting system (mp > se > be > ic > default) with configurable edition preferences
  • Support for float version comparison (e.g., 19.5 > 19) within same edition type
  • Automatic path detection and version parsing from executable names and paths
  • Comprehensive docstring with practical examples for easy integration
  • Fallback priority handling for unknown editions with graceful degradation

• Project Organization Improvements: Enhanced directory structure for better maintainability

  • Moved agent_examples directory to source/agent_examples for centralized documentation
  • Updated all documentation references to reflect new directory structure
  • Maintained file integrity through git rename operations preserving history
  • Improved project structure consistency with better separation of concerns

Changed

• Documentation Synchronization: Ensured consistency across multilingual documentation

  • Updated Chinese README news section with English version content including StataFinder updates
  • Added newest research section with links to reports website for better user guidance
  • Enhanced environment variable configuration documentation with setup examples
  • Maintained consistency between English and Chinese documentation for better user experience

• StataFinder Architecture: Enhanced cross-platform support with Linux integration

  • Enabled Linux platform in FINDER_MAPPING removing "Not Support now" limitation
  • Improved error handling with proper exception messages for missing Stata installations
  • Enhanced type annotations and documentation throughout the StataFinder system
  • Maintained backward compatibility while adding comprehensive Linux support

Fixed

• Documentation Consistency: Resolved discrepancies between English and Chinese README files
  • Added missing StataFinder update notices with environment variable configuration guidance
  • Included newest research links that were present in English but missing in Chinese version
  • Ensured all news content is properly synchronized across supported languages
  • Improved documentation accuracy for better user guidance and setup experience

Technical

• Code Quality: Enhanced type safety and error handling throughout Linux finder implementation
• Performance: Optimized directory search with intelligent path validation and iteration
• Compatibility: Maintained full backward compatibility with existing StataFinder functionality
• Extensibility: Clean architecture allows easy addition of new platforms and search strategies
• Documentation: Comprehensive docstrings and examples for enhanced developer experience

Click to expand v1.12.0 details

[1.12.0] - 2025-10-31

Added

• Unified Priority System: Complete StataFinder architecture overhaul with intelligent version and edition comparison

  • New StataEditionConfig class with comprehensive version extraction and sorting capabilities
  • Priority system: MP > SE > BE > IC > default with configurable edition preferences
  • Automatic version detection from file paths, directory names, and executable names
  • Support for floating-point versions (e.g., 19.5 > 19) and reasonable version range validation (1-30)
  • Magic methods for easy data access (__str__, __int__, __float__) and path properties

• Complete Windows Platform Support: Full implementation of FinderWindows with three-layer search strategy

  • Layer 1: Standard installation path search (Program Files, Program Files (x86))
  • Layer 2: Non-standard custom installation locations with drive scanning
  • Layer 3: Deep scan fallback for comprehensive discovery across all drives
  • Intelligent drive detection and depth-limited scanning for performance optimization
  • Skip pattern system to avoid redundant searching of standard directories in non-standard scans

• Enhanced macOS Application Detection: Improved FinderMacOS with comprehensive app bundle support

  • /Applications directory scanning with version extraction from isstata.* files
  • Automatic edition detection from Stata app names (StataMP, StataSE, StataBE, StataIC)
  • Support for both /Applications/Stata unified directory and separate versioned applications
  • Robust executable path resolution within app bundles (Contents/MacOS/stata-{edition})

Fixed

• Critical Empty List Handling: Prevents crashes when no Stata installations are found

  • Fixed max() calls on empty lists in both Windows and macOS finder methods
  • Proper empty list validation before attempting to find highest priority Stata executable
  • Graceful fallback behavior when search layers return no results
  • Enhanced error handling with meaningful exception messages

• Platform-Specific Search Optimization: Improved search efficiency and reliability

  • Windows: Simple matching approach replaces overly strict regex filtering for broader compatibility
  • macOS: Better handling of Homebrew installations without version numbers (defaults to version 99)
  • Enhanced path validation and executable checking across all platforms
  • Improved exception handling for permission errors and inaccessible directories

Changed

• Architecture Modernization: Complete StataFinder system redesign for better maintainability

  • Abstract FinderBase class with comprehensive documentation and type annotations
  • Platform-specific implementations following consistent design patterns
  • Enhanced finder.py integration with improved error handling and exception management
  • Updated utility modules (usable.py, installer.py, mcp_servers.py) for better integration

• Enhanced Documentation and Type Safety: Improved code quality and developer experience

  • Comprehensive docstrings for all new classes and methods with examples
  • Full type annotations throughout the StataFinder system
  • Detailed abstract method documentation explaining platform-specific requirements
  • Clear separation of concerns between different search strategies and platform implementations

Technical

• Performance: Three-layer search strategy balances speed with thoroughness
• Reliability: Robust error handling prevents crashes across all search scenarios
• Compatibility: Maintains backward compatibility while adding new functionality
• Extensibility: Clean architecture allows easy addition of new platforms and search strategies
• Testing: Comprehensive validation across different installation patterns and platforms

Click to expand v1.11.1 details

[1.11.1] - 2025-10-23

Fixed

• Cross-Platform Initialization Error: Critical fix for Windows and Linux StataFinder initialization failure

  • Fixed AttributeError in src/stata_mcp/core/stata/stata_finder/finder.py:43
  • Updated finder initialization from finder_mapping.get(self.current_os)() to _os_finders.get(self.current_os, None)
  • Resolves RuntimeError on Windows and Linux systems during StataFinder initialization
  • Ensures consistent behavior across all supported platforms (macOS, Windows, Linux)

• Platform Compatibility: Enhanced cross-platform stability and reliability

  • Proper None handling in StataFinder initialization prevents function call errors
  • Eliminates startup failures for Windows and Linux users
  • Maintains existing functionality for macOS users
  • Improved error prevention with defensive programming practices

Technical

• Code Quality: Single-line critical fix with maximum impact and minimal risk
• Testing: Comprehensive cross-platform verification completed
• Compatibility: 100% backward compatible with existing installations
• Stability: Enhanced system reliability for all supported operating systems

Click to expand v1.11.0 details

[1.11.0] - 2025-10-20

Added

• Claude Code Client Mode: Added dedicated support for Claude Code client environment
  • New -c/--client CLI option for Claude Code client mode
  • Automatic detection and configuration for Claude Code environment
  • Enhanced working directory handling for client-specific workflows
  • Optimized MCP server integration for Claude Code platform

Changed

• Code Architecture: Major refactoring to improve modularity and maintainability
  • Extracted MCP server initialization and tools from __init__.py to dedicated mcp_servers.py module
  • Updated main.py import path to use new modular structure
  • Streamlined package initialization with better separation of concerns
  • Improved code organization for enhanced development experience

Changed

• Dependency Management: Optimized dependency structure for cleaner installation
  • Removed python-dotenv from active dependencies as it's no longer required
  • Updated uv.lock file to reflect dependency changes
  • Improved CLI import structure for better performance
  • Enhanced version import using importlib.metadata instead of package-level import

Changed

• CLI Enhancements: Improved command-line interface with new client mode support
  • Restructured CLI imports for better loading performance
  • Added lazy loading patterns for non-critical components
  • Enhanced error handling and user feedback
  • Better separation between different CLI modes (agent, client, mcp)

Technical

• Performance: Enhanced startup performance through optimized imports and modular architecture
• Maintainability: Improved code organization with clear separation between server initialization and package metadata
• Compatibility: Better support for different client environments and use cases
• Dependencies: Streamlined dependency list with removal of unnecessary python-dotenv requirement

Click to expand v1.10.2 details

[1.10.2] - 2025-10-17

Added

• AgentRunner Evaluation Method: Complete evaluation workflow implementation in AgentRunner class

  • New evaluate() method for seamless ScoreModel integration with automatic task execution and scoring
  • Type conversion handling for processer data (List[Dict] to String format) for ScoreModel compatibility
  • Optional result display with is_display parameter for immediate feedback and debugging
  • Automatic conversation history extraction and formatting for evaluation purposes
  • Direct parameter passing and type-safe integration between AgentRunner and ScoreModel

• Enhanced Evaluation Integration: Streamlined ScoreModel connection with agent execution

  • Automatic task execution and result processing in a single method call
  • Built-in error handling and validation throughout the evaluation pipeline
  • Comprehensive fallback mechanisms for robust evaluation operations
  • Simplified evaluation workflow with reduced boilerplate code

Changed

• Developer Experience: Improved evaluation workflow simplicity and usability

  • Single evaluate() call replaces multi-step evaluation process
  • Automatic handling of type conversions between AgentRunner and ScoreModel components
  • Optional display functionality for real-time monitoring and debugging
  • Enhanced error reporting and validation for evaluation scenarios

• Type Compatibility: Enhanced data flow between evaluation components

  • Robust integration between different data structures (List[Dict] to String conversion)
  • Preserved conversation structure and context during type conversion
  • Maintained data integrity throughout the evaluation pipeline
  • Type-safe parameter handling for reliable evaluation workflows

Technical

• Code Architecture: Enhanced AgentRunner with complete evaluation capabilities

  • Clean separation between task execution and evaluation logic
  • Modular design allowing for easy customization and extension
  • Comprehensive type annotations and documentation for evaluation methods
  • Improved error handling and validation throughout the evaluation framework

• Backward Compatibility: Maintained full compatibility with existing evaluation workflows

  • All existing AgentRunner functionality remains unchanged
  • New evaluate() method is additive without disrupting existing code
  • Seamless integration with existing ScoreModel evaluation framework
  • Zero impact on current evaluation processes and tools

Click to expand v1.10.1 details

[1.10.1] - 2025-10-16

Added

• AgentRunner Class: New comprehensive agent management system for simplified LLM evaluation workflows

  • OpenAI-compatible API integration with support for multiple providers (OpenAI, DeepSeek, etc.)
  • get_processer() static method for automatic conversation history extraction from agent results
  • get_final_result() static method for extracting final answers from agent outputs
  • Built-in error handling and graceful fallback mechanisms for robust operation
  • Full type annotations and comprehensive documentation with practical examples

• Comprehensive Testing Framework: Complete test suite for AgentRunner functionality validation

  • Mock-based testing system for reliable validation without API dependencies
  • Real API testing with automatic fallback to mock tests when credentials unavailable
  • Comprehensive error handling validation and conversation structure verification
  • Integrity checks for method outputs and data format validation
  • Complete test coverage in tar.py with both mock and real API scenarios

• Enhanced Documentation: Updated evaluation guide with AgentRunner integration examples

  • Step-by-step AgentRunner usage instructions with practical Stata analysis examples
  • Complete code examples for real-world evaluation scenarios including auto dataset analysis
  • Batch evaluation workflows for processing multiple tasks efficiently
  • Custom evaluation criteria and metrics implementation examples
  • Environment setup guides for different API providers (OpenAI, DeepSeek)

Changed

• Module Organization: Enhanced evaluate module structure with better component integration

  • AgentRunner now properly exported in evaluate.__init__.py for direct imports
  • Cleaner module structure with improved separation of concerns
  • Better discoverability of evaluation components and enhanced import experience
  • Updated Evaluation.md documentation with comprehensive AgentRunner usage examples

• Developer Experience: Streamlined evaluation workflow with simplified task execution

  • One-click conversation extraction and result processing through AgentRunner methods
  • Reduced boilerplate code for common evaluation scenarios and use cases
  • Better error messages and debugging information for improved development experience
  • Enhanced integration between agents and evaluation tools

• Evaluation Framework: Improved evaluation workflow with better tool integration

  • Seamless integration between AgentRunner and existing ScoreModel functionality
  • Enhanced support for OpenAI Agents result processing and extraction
  • Better handling of complex conversation structures with multi-turn interactions
  • Improved role detection and content extraction for diverse agent response formats

Fixed

• Conversation Processing: Enhanced handling of complex conversation structures in agent outputs

  • Better support for multi-turn conversations with tool interactions and role switching
  • Enhanced role detection and content extraction from various agent response formats
  • More robust parsing of agent response data with improved error resilience
  • Better handling of edge cases in conversation structure processing

• Testing Reliability: Enhanced test stability and comprehensive error handling

  • Improved mock data structure for more realistic testing scenarios
  • Better exception handling in test environments with detailed error reporting
  • More comprehensive validation of method outputs and data integrity
  • Enhanced test reliability through better error recovery mechanisms

Technical

• Dependencies: No new dependencies added for AgentRunner functionality

  • AgentRunner uses existing project dependencies (openai-agents, openai, langchain)
  • Lightweight implementation with minimal performance impact
  • Backward compatibility maintained for all existing features and workflows

• Code Architecture: Enhanced evaluation module with improved component integration

  • Clean separation between agent execution and result processing logic
  • Modular design allows for easy extension and customization
  • Standardized evaluation criteria based on professional statistical standards
  • Improved error handling and fallback mechanisms throughout the framework

Click to expand v1.10.0 details

[1.10.0] - 2025-10-14

Added

• LLM Evaluation Module Framework: Comprehensive evaluation system for Large Language Model performance assessment

  • New evaluate package with complete module structure (src/stata_mcp/evaluate/)
  • _model.py providing structured assessment framework and evaluation criteria
  • advice.py for evaluation advice generation and result analysis
  • score_it.py for automated LLM performance scoring and assessment
  • Full type annotations and comprehensive documentation for evaluation use

• ScoreModel Evaluation System: Automated LLM performance assessment framework

  • Task completion accuracy evaluation based on reference answers
  • Response quality assessment against expected outcomes
  • Process analysis for LLM reasoning quality evaluation
  • Historical message processing for context assessment
  • Quantitative metrics for model comparison and benchmarking

• Structured Evaluation Framework: Systematic approach to LLM performance evaluation

  • Task definition and reference answer management system
  • Process evaluation for analyzing LLM reasoning chains
  • Final answer scoring and validation mechanisms
  • Configurable evaluation parameters for different use cases
  • Extensible framework for custom evaluation criteria

• Research Assessment Tools: Built-in evaluation capabilities for AI research workflows

  • Standardized evaluation methodology for reproducible research
  • Benchmarking framework for model performance comparison
  • Reference answer management for accuracy assessment
  • Process analysis capabilities for reasoning evaluation

Changed

• Enhanced Research Capabilities: Integrated LLM evaluation into existing Stata-MCP functionality

  • Evaluation framework works seamlessly with current research workflows
  • Optional enhancement for AI performance assessment without disrupting existing functionality
  • Flexible configuration for different evaluation scenarios and research requirements
  • Multi-language support for evaluation feedback and assessment results

• Documentation Updates: Enhanced user documentation with LLM evaluation features

  • Updated main README with LLM evaluation module announcement
  • Synchronized Chinese README with new evaluation capabilities
  • Added Evaluation.md documentation for LLM assessment workflows
  • Enhanced quick start guides with evaluation setup examples

Technical

• Code Architecture: New evaluation module with clean separation of concerns

  • Modular design allows for easy extension and customization
  • Standardized evaluation criteria based on professional statistical standards
  • Configurable scoring parameters for different assessment requirements
  • Professional statistical standards integration for academic use

• Dependencies: No new dependencies added for evaluation functionality

  • Evaluation module uses existing project dependencies
  • Lightweight implementation with minimal performance impact
  • Backward compatibility maintained for all existing features

Click to expand v1.9.1 details

[1.9.1] - 2025-10-12

Removed

• WebUI Module: Completely removed Flask-based web interface and all related functionality

  • Removed Flask dependency and all web-related dependencies from pyproject.toml
  • Deleted webui module (src/stata_mcp/webui/) including templates, static assets, and utilities
  • Removed --webui CLI argument and related webui startup functionality
  • Deleted WEBUI.md documentation file

• Config Module: Discontinued TOML-based configuration system

  • Removed config module (src/stata_mcp/config/) and all configuration management logic
  • Deleted Configuration.md documentation file
  • Removed example.toml configuration template
  • Simplified project to use environment variable-based configuration only

Changed

• Project Structure: Major simplification to focus on core MCP functionality

  • Streamlined project structure by removing 17 files and 2299 lines of code
  • Reduced dependency footprint for faster installation and smaller package size
  • Improved startup performance by eliminating complex configuration loading
  • Enhanced code maintainability through reduced complexity

• Dependency Management: Optimized dependency list for lightweight installation

  • Removed Flask and web-related dependencies
  • Commented out optional jupyter-related dependencies not currently used
  • Maintained all core dependencies for Stata-MCP functionality
  • Updated uv.lock to reflect new dependency structure

• Configuration Simplification: Streamlined configuration approach

  • Now relies solely on environment variables for configuration
  • Maintained dotenv loading for essential settings
  • Removed complex TOML parsing and validation logic
  • Better defaults for improved out-of-the-box experience

Technical

• Code Quality: Improved maintainability and focus

  • Reduced attack surface by removing web interface components
  • Fewer potential points of failure in initialization
  • Cleaner separation of concerns between CLI and core functionality
  • Better alignment with MCP protocol's primary use case

• Performance: Enhanced startup and runtime performance

  • Faster initialization due to reduced module loading
  • Lower memory footprint from fewer loaded dependencies
  • Simplified error handling paths
  • Improved reliability through reduced complexity

Migration Notes

• WebUI Users: Web interface no longer available - use CLI interface instead
• Config File Users: TOML configuration no longer supported - migrate to environment variables
• Minimal Impact: Most users unaffected as they were already using CLI-based workflow
• Simple Migration: Straightforward migration path for affected users

Click to expand v1.9.0 details

[1.9.0] - 2025-10-11

Added

• Agent as Tool Framework: Revolutionary multi-agent workflow support for Stata analysis

  • New StataAgent class with comprehensive ReAct (Reasoning-Action-Observation) framework
  • Professional Stata Data Analysis Expert role with economic research assistant capabilities
  • Seamless integration as a tool within other AI agents for complex workflows
  • Default comprehensive instructions covering data understanding, code generation, execution, and results interpretation
  • Configurable tool descriptions with clear capabilities and input/output specifications

• Multi-Model Provider Support: Enhanced compatibility with various LLM providers

  • Extended type hints to support OpenAIChatCompletionsModel | Model union types
  • set_model() utility function for easy configuration of alternative providers (DeepSeek, etc.)
  • Native support for OpenAI ChatCompletionsModel with fallback to generic Model interface
  • Improved IDE support and code completion through enhanced type annotations

• Comprehensive Documentation Suite: Complete bilingual documentation for Agent as Tool functionality

  • Detailed agent_as_tool.md guide with practical examples and use cases
  • Quick start guide for basic usage and advanced configuration examples
  • Integration patterns with existing agent frameworks (OpenAI Agents, LangChain)
  • Multi-provider setup examples with DeepSeek and other OpenAI-compatible models

• Enhanced README Integration: Improved user onboarding with Agent as Tool examples

  • New "Agent as Tool" section in main README with working code examples
  • Synchronized Chinese README with all new features and examples
  • Updated news sections highlighting new multi-agent capabilities
  • Clear navigation to detailed documentation and quick start guides

Changed

• Agent Architecture Enhancements: Modular and flexible agent design

  • Clean separation between agent logic and tool integration through as_tool property
  • Flexible MCP server configuration with environment variable support
  • Enhanced tracing control for performance optimization and debugging
  • Configurable agent behavior through custom instructions, models, and tools

• Developer Experience Improvements: Enhanced usability and flexibility

  • Support for custom agent names, instructions, and tool descriptions
  • Adjustable max_turns parameter for complex analysis tasks
  • Configurable DISABLE_TRACING for performance optimization
  • Better error handling and connection management for MCP servers

• Type Safety and IDE Support: Improved development experience

  • Enhanced type annotations throughout the agent framework
  • Support for union types (OpenAIChatCompletionsModel | Model)
  • Better IDE support with improved code completion and error detection
  • Reduced runtime errors through comprehensive type checking

Technical

• New Dependencies: Added agents library for Agent as Tool functionality

  • Support for OpenAI Agents framework with seamless integration
  • Enhanced MCP server configuration and management
  • Improved error handling and connection stability

• Code Organization: Enhanced module structure for agent functionality

  • New agent_as_tool module with StataAgent and set_model utilities
  • Clean separation of concerns between agent logic and tool integration
  • Improved maintainability and extensibility for future agent features

• Configuration Flexibility: Enhanced customization options

  • Environment variable support for Stata CLI configuration
  • Customizable agent instructions and tool descriptions
  • Flexible model provider configuration with easy switching between providers

Click to expand v1.8.2 details

[1.8.2] - 2025-10-10

Added

• URL-based DTA File Reading: Implemented support for reading Stata DTA files directly from HTTP/HTTPS URLs
  • Added URL detection and validation in DtaDataInfo class
  • Implemented memory-efficient reading using BytesIO for optimal performance
  • Support for both http:// and https:// protocols with comprehensive error handling
  • Seamless integration with existing local file functionality
  • Automatic URL format validation and file extension verification

Changed

• Enhanced DtaDataInfo Architecture: Extended DtaDataInfo class to support both local and remote files

  • Unified _read_data() method handles both file sources without code duplication
  • Intelligent file path detection (URL vs local path) for appropriate processing
  • Streamlined DataInfoBase class by removing unnecessary abstract method _read_data_from_url()
  • Cleaner separation of concerns between local and remote operations

• Documentation Updates: Enhanced API documentation and examples

  • Updated docstring examples in __init__.py with comprehensive output format examples
  • Added clear demonstration of remote data analysis capabilities
  • Improved error handling documentation for network operations

Fixed

• URL Handling Conflicts: Resolved Path object conflicts when processing URLs as file paths

  • Fixed URL validation to properly parse and validate HTTP/HTTPS URLs
  • Corrected file extension checking for remote files using URL path parsing
  • Improved error messages to provide clearer feedback for URL-related issues

• Network Error Handling: Enhanced error handling for network operations

  • Added comprehensive timeout configurations and status code validation
  • Improved error reporting for network failures, invalid URLs, and file format issues
  • Better exception handling with informative error messages for troubleshooting

Technical

• New Dependencies: Added requests library for HTTP operations

  • Robust HTTP client with built-in error handling and retry mechanisms
  • Efficient content handling for binary DTA files with proper encoding
  • Support for HTTP/HTTPS protocols with TLS security

• Memory Optimization: Implemented memory-efficient data processing

  • Direct loading into BytesIO eliminates temporary file overhead
  • Streaming content handling for large files with optimal memory usage
  • Zero intermediate file operations for URL-based data access

Click to expand v1.8.1 details

[1.8.1] - 2025-10-09

Fixed

• MCP Version Compatibility: Enhanced MCP initialization logic for better version compatibility
  • Improved error handling for different MCP implementations (v1.16.0+)
  • Enhanced FastMCP initialization with fallback mechanisms
  • Better robustness when encountering validation errors
  • Streamlined version-specific initialization logic

Changed

• Initialization Process: Reorganized FastMCP initialization sequence
  • Prioritized MCP v1.16.0+ configuration with proper icon array format
  • Improved fallback error handling with non-config initialization
  • Enhanced user guidance for MCP version upgrades

Technical

• Code Robustness: Improved error handling and initialization reliability
• Version Support: Better compatibility with MCP v1.16.0 and newer versions

Click to expand v1.8.0 details

[1.8.0] - 2025-10-09

Added

• Enhanced Data Info Support: Added comprehensive multi-format data information functionality

  • New CsvDataInfo class for handling CSV file metadata and statistics
  • Enhanced DtaDataInfo class for Stata .dta file data information extraction
  • Improved DataInfoBase base class with kwargs support for extensibility
  • Enhanced get_data_info function with CSV and enhanced DTA file format support
  • Added save functionality with configurable output options
  • Implemented temporary directory management for data processing
  • Full type annotations and comprehensive documentation

• Enabled Data Info Tool: Reactivated get_data_info functionality

  • Tool decorator re-enabled for production use
  • Support for multiple file formats: .dta, .csv, and Excel files
  • Improved error handling and user feedback
  • Enhanced data summary statistics and metadata extraction

Changed

• MCP Version Upgrade: Updated MCP dependency from v1.15.0 to v1.16.0
  • Enhanced FastMCP initialization logic for better version compatibility
  • Improved error handling for different MCP implementations
  • Better robustness in initialization process
  • Streamlined dependency management

Technical

• Code Architecture: Improved data info module structure with base classes
• Type Safety: Enhanced type annotations across data info functionality
• Module Organization: Better separation of concerns in data processing modules

Click to expand v1.7.4 details

[1.7.4] - 2025-10-09

Fixed

• MCP Dependency Conflicts: Resolved version conflicts in MCP dependency chain
  • Fixed compatibility issues with different MCP implementations
  • Enhanced system stability through streamlined dependency management
  • Improved error handling for dependency-related edge cases

Technical

• Dependency Optimization: Streamlined MCP dependencies for better stability
• Version Update: Updated version from 1.7.3 to 1.7.4

Click to expand v1.7.3 details

[1.7.3] - 2025-10-06

Added

• Enhanced Encoding Support: Added configurable encoding parameters to dofile functions
  • write_dofile now supports optional encoding parameter (default: utf-8)
  • append_dofile enhanced with configurable encoding for read/write operations
  • Better support for international character sets including Chinese, Japanese, Korean
  • Maintains full backward compatibility with existing code

Fixed

• Issue #18: Resolved potential Chinese character encoding problems in dofile operations
  • Proactive fix for international character set support
  • Enables flexible encoding handling for various environments

Technical

• Encoding Flexibility: Improved dofile encoding handling without breaking changes
• Version Update: Updated version from 1.7.2 to 1.7.3

Click to expand v1.7.2 details

[1.7.2] - 2025-10-06

Added

• Agent Mode Support: Added comprehensive agent mode functionality
  • New stata-mcp --agent command line option for interactive AI-driven analysis
  • StataAgent class with LangChain integration for AI-powered Stata operations
  • Support for GPT-5, DeepSeek, and other OpenAI-compatible models
  • ReAct (Reasoning + Acting) prompt template for enhanced AI reasoning
  • MultiServerMCPClient integration for seamless Stata command execution
  • Interactive workflow with data source and task input prompts
• Agent Examples: Added complete agent implementation examples
  • LangChain and LangGraph integration examples
  • OpenAI-based agent implementation
  • Advanced prompt generation system for agent tasks
  • Comprehensive documentation and README files
• Agent Startup Script: Added agent.sh automated startup script
  • Auto-detection of uv package manager with fallback to pip
  • Python 3.11+ version validation and environment checking
  • Interactive installation prompts for uv package manager
  • Automatic package installation and version validation
  • Seamless agent mode launch with proper error handling
  • Cross-platform compatibility with colored output for better UX

Changed

• Jupyter Dependencies: Removed unused Jupyter-related dependencies from pyproject.toml
  • Commented out jupyter-client>=8.6.3 and stata-kernel>=1.12.2
  • Commented out notebook>=7.4.5 and jupyter>=1.1.1
  • Streamlined installation process and reduced package size
  • No impact on core Stata-MCP functionality

Technical

• Dependency Optimization: Cleaned up unused dependencies for faster installation
• Agent Mode Integration: Enhanced CLI with agent mode support via -a/--agent flag
• Version Update: Updated version from 1.7.1 to 1.7.2

Click to expand v1.7.1 details

[1.7.1] - 2025-10-05

Changed

• mk_dir Security: Re-enabled mk_dir tool with enhanced security using pathvalidate library
  • Added comprehensive path validation and sanitization
  • Implemented secure directory creation with proper permissions (0o755)
  • Added detailed error handling for invalid paths and permission issues
  • Improved function documentation with comprehensive parameter descriptions

Technical

• Dependencies: Added pathvalidate>=3.3.1 for secure path validation
• Version Update: Updated version from 1.7.0 to 1.7.1

Click to expand v1.7.0 details

[1.7.0] - 2025-10-4

Added

• AI-Assisted Research: Added comprehensive AI-assisted empirical research report with latest findings
• Prompt Engineering: Introduced comprehensive task prompt guide and examples for better AI interaction
• Template System: Added prompt-generator template for standardized AI request formatting
• Research Documentation: Added detailed research report on StataMCP usage for social science research
• Prompt Examples: Added two practical examples for PromptGenerator usage
• Load Figure: Added load_figure functionality for image handling
• Sandbox Infrastructure: Added sandbox infrastructure for testing
• Main Entry Point: Added main.py entry point for local development
• Multilingual Documentation: Updated Chinese, French, and Spanish README files
• Agent Mode Support: Added agent mode support documentation
• China Users Guide: Added specialized documentation for China users
• LLM Integration Guide: Added comprehensive LLM documentation for AI integration

Changed

• MCP Protocol: Upgraded MCP from version 1.14 to 1.15 for latest features
• Module Naming: Renamed StataFinder to stata_finder for snake_case consistency
• Directory Structure: Improved code organization with better directory structure
• Contributing Guide: Updated CONTRIBUTING.md documentation
• Security Policy: Enhanced security policy with comprehensive privacy disclaimer
• Git Standards: Updated CLAUDE.md with git push restrictions and standards
• App Icon: Updated app icon to higher resolution image
• Project Dependencies: Updated project dependencies and lock files

Disabled (Commented Out)

• Directory Creation: mk_dir tool implemented but decorator commented out for safety considerations
• Data Info Function: get_data_info tool implemented but decorator commented out

Fixed

• ValueError Prevention: Fixed default instructions setting to avoid ValueError
• macOS Compatibility: Fixed errors in StataFinder.macos
• Citation Corrections: Fixed citation mistakes in research documentation
• Environment Configuration: Fixed environment name errors in configuration
• Debug Cleanup: Removed leftover debug print statements

Technical

• Code Architecture: Improved code organization and maintainability
• Function Safety: Temporarily disabled certain functions via decorator commenting for security
• Development Environment: Enhanced development environment setup
• Data Info Stability: Temporarily disabled get_data_info for stability considerations

Click to expand v1.6.3 details

[1.6.3] - 2025-09-12

Added

• MCP Resource Support: Added @mcp.resource decorator for help function with URI help://stata/{cmd}
• AI Coding Ability Report: Added comprehensive comparison chart of different AI models' Stata code generation capabilities in source documentation

Changed

• Dependency Upgrade: Upgraded mcp[cli] from >=1.9.0 to >=1.13.0 for latest MCP protocol features
• Version Updates: Updated version numbers across all documentation files and CITATION.cff

Technical

• Enhanced MCP Support: Improved MCP protocol compatibility with resource URI support

Click to expand v1.6.2 details

[1.6.2] - 2025-08-15

Changed

• CLI Architecture: Refactored CLI entry point from __init__.py to dedicated CLI module
  • Moved CLI functionality to stata_mcp/cli/_cli.py
  • Improved code modularity and separation of concerns
  • Updated entry point configuration in pyproject.toml
  • Enhanced maintainability following Python packaging best practices

Technical

• Code Organization: Clean separation between package initialization and CLI execution
• Entry Point: Updated to use dedicated CLI module instead of __init__.py
• Module Structure: Reduced complexity in main module initialization

Click to expand v1.6.1 details

[1.6.1] - 2025-08-09

Fixed

• Fixed Excel file reading issue in get_data_info function by adding missing openpyxl dependency
• Resolved compatibility issues with Excel (.xlsx) file formats

Added

• Added openpyxl>=3.1.5 to project dependencies for Excel file support

Security

• Updated license from MIT to Apache License 2.0 for better legal protection and compatibility

Changed

• License: Migrated from MIT License to Apache License 2.0
  • Updated LICENSE file to Apache 2.0 full text
  • Updated all documentation files (README, README-cn, README-fr, README-sp)
  • Updated CITATION.cff license field
  • Updated pyproject.toml license field
  • Updated Statement.md in all languages (中文, English, Français)
  • Updated all license badges from MIT to Apache 2.0
  • Ensured consistent Apache 2.0 licensing across entire project

Click to expand v1.6.0 details

[1.6.0] - 2025-06-28

Added

• Initial release with core Stata-MCP functionality
• Support for regression analysis via LLM integration
• Multi-language documentation (English, Chinese, French, Spanish)
• PyPI package distribution
• Jupyter integration support
• Web UI interface
• Cross-platform support (macOS, Windows, Linux)

Features

• Stata command execution via MCP protocol
• Data analysis automation
• Regression model building assistance
• Statistical output interpretation
• Code generation and debugging support