Skip to content

mengshi02/axons

Axons

axons Logo

The Ultra-Lightweight AI-First Code Workbench

Lightweight by Design · Extensible by Nature · Instant Startup · Native AI Intelligence · Private Deployment

CI Release License Go Version Issues


Axons is an ultra-lightweight AI-First code workbench — self-built four-dimensional intelligent engine core, AI capabilities natively embedded not bolted on. 5+ AI expert agents ready out-of-the-box, covering architecture governance, tech debt cleanup, and legacy system iteration. Your code never leaves your machine, fully private end-to-end.

Website: axons.chat

Core Capabilities

Ultra-Lightweight & Efficient

Ditch bloated traditional IDE features. Instant startup, minimal memory footprint, zero lag. Even low-spec devices run smoothly through the entire development cycle.

Self-Built 4D Intelligent Engine

Graph Computing Engine (GCE), Analysis Engine (ACE), Cognitive Context Engine (CCE), and LLM — four engines deeply fused into a native AI code intelligence foundation, delivering full-dimension code perception and precise context understanding.

Native Expert-Level AI Coding

5+ vertical-domain AI expert agents built-in, ready out-of-the-box with zero configuration. Covering code writing, bug fixing, architecture optimization, tech debt cleanup, and documentation generation. AI capabilities natively empowered, not simple plugin adaptations.

  • AI Orchestrator — intelligent task decomposition and delegation, multi-expert coordination

  • Architect — identifies module boundaries, analyzes dependencies, checks architecture compliance

  • Quality Analyst — detects code smells, dead code, hotspot functions, excessive coupling

  • Impact Analyst — evaluates change impact scope and blast radius, aids Code Review decisions

  • Code Engineer — read/write files, execute commands, end-to-end coding tasks, integrated native terminal

  • Custom Agent — create domain-specific agents with one click, persistent memory

  • AI Coding Safety: precise scope modification based on graph dependencies, rejects large-scale brute rewrites

  • Operation rollback: every AI edit is atomically recorded, supporting line-level precise rollback

Open Extension Ecosystem

Plugin system covering development languages, frameworks, tools, MCP protocol, custom Skills, and enterprise components. Highly customizable for individuals and teams. Assemble on demand, scale flexibly — zero resource usage when not enabled.

  • Full plugin lifecycle: install, activate, deactivate, uninstall with manifest protocol
  • Frontend plugin SDK with iframe isolation, component library, and theme support
  • Backend plugin support with health check, environment variables, and platform overrides
  • Lazy loading via activation events — plugins only load when needed
  • Official extension repository: axons-extension-packages

Full-Scenario Remote Development

Native compatibility with Docker, WSL, and SSH remote development. Remote environment feels identical to local. The local side stays ultra-lightweight, perfectly adapting to cloud-native, distributed, and remote collaboration scenarios.

Open Source Without Limits

MIT license, fully open source, no paywalls, no feature cuts, no commercial restrictions. Supports secondary development, custom modifications, and internal enterprise deployment.

Incremental Graph Dynamic Update — Core Performance Edge

Only detects changed files, functions, and dependencies — recomputes only the delta, never the entire project graph.

  • ⚡ Sub-second architecture graph refresh, million-line projects update seamlessly
  • 💾 Background resident memory footprint ultra-low, virtually imperceptible
  • 🔄 Intelligent cascade detection: Journal → mtime → content hash
  • 🗄️ Multi-project isolated storage, cache auto-reuse

Supported Languages

Go, C, C++, Java, Python, Rust, C#, JavaScript, TypeScript

Installation

Download

Download the latest release from GitHub Releases:

  • macOS: Apple Silicon / Intel
  • Windows: x86_64 / ARM64
  • Linux: x86_64 / ARM64 (AppImage / DEB / RPM)
  • Web: cross-platform browser access

Build from Source

git clone https://github.com/mengshi02/axons.git
cd axons
make build

Quick Start

# Start the daemon
./axons daemon start --tcp :8080

# Access the Web UI
open http://localhost:8080

# Build code graph
./axons build /path/to/your/code

# Search code
./axons search "functionName"

CLI Commands

Command Description
build Build code graph (with incremental updates)
query Query nodes, edges, and relationships
search Multi-mode search (keyword/semantic/regex)
audit Code quality audit and cycle detection
complexity Code complexity analysis
owners Map code ownership from git history
path Find paths between symbols
sequence Generate call sequence diagrams
cfg Generate control flow graphs
dataflow Data flow analysis
cochange Identify co-changing files
diff-impact Analyze impact of git diffs
branch-compare Compare branches for differences
snapshot Create and manage code snapshots
watch Real-time file monitoring and graph updates
embed Generate embeddings for semantic search
export Export graph data
registry Manage multiple projects
stats Project statistics
triage Triage issues and locate affected code

Agent Skills

Three built-in agent skills following the Agent Skills Specification:

Skill Description
code-graph-analyzer Code architecture and quality analysis
dependency-tracker Dependency tracing and circular dependency detection
code-search-assistant Intelligent code search (multi-mode support)

See skills/README.md for details.

MCP Integration

Axons provides an MCP (Model Context Protocol) server exposing 30+ tools for AI assistants, including:

  • Keyword search, semantic search, hybrid search
  • Symbol lookup, call chain analysis, path finding
  • Impact analysis, complexity metrics, PageRank
  • Architecture compliance checking, community detection
  • File system operations and command execution
  • Cognitive context engine (bimodal embedding + scenario templates)
  • Agent delegation (multi-agent collaboration)

LLM Support

  • OpenAI (GPT-4, GPT-3.5, etc.)
  • Anthropic (Claude)
  • Custom LLM endpoints

Embedding Providers

  • OpenAI (text-embedding-3-small/large)
  • Jina AI
  • Custom embedding endpoints

Development

Prerequisites

  • Go 1.25+
  • Node.js 22+

Build

make deps          # Install dependencies
make build         # Build binary
make test          # Run tests
make lint          # Run linter

Development Mode

make frontend-dev  # Start frontend dev server
make daemon        # Start backend in another terminal

Desktop App

The project provides a desktop client built with Electron:

make desktop-dev           # Development mode
make desktop-build         # Build desktop app
make desktop-build-mac     # Build for macOS
make desktop-build-windows # Build for Windows

Documentation

Contributing

Contributions are welcome. See CONTRIBUTING.md for guidelines.

License

MIT License. See LICENSE for details.

Contact

About

An ultra-lightweight AI-First code workbench

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors