Skip to content

FRAMEWORK.md

Latest commit

 

History

History
147 lines (104 loc) · 3.54 KB

FRAMEWORK.md

File metadata and controls

147 lines (104 loc) · 3.54 KB
Raw

Loop Engineering MCP Server - Implementation Framework

The easiest way to implement loop engineering in your AI coding workflow.

This MCP (Model Context Protocol) server runs locally on your machine and integrates directly with Cursor, Kiro, Claude Desktop, or any MCP-compatible AI agent. No hosting required, no containers to manage.

What This MCP Server Does

Adds powerful loop engineering tools to your AI agent:

  • create_loop - Set up automated coding loops (CI triage, dependency updates, lint fixes)
  • start_loop / stop_loop - Control loop execution
  • list_loops - View all active and paused loops
  • add_skill - Create reusable skill templates
  • view_state - Check loop history and metrics
  • configure_verification - Set up automated testing gates

Your AI agent uses these tools automatically when you say things like:

  • "Set up a CI triage loop that runs every 6 hours"
  • "What loops are currently running?"
  • "Create a dependency update loop"

Prerequisites

Before implementing any loop, verify you pass the 4-Condition Test:

  • Task repeats at least weekly
  • Verification is automated (tests, linter, type checker, build)
  • Token budget can absorb retry costs
  • Agent has senior tools (can run code, see logs, reproduction environment)

If you fail any condition, stop here and use manual prompting instead.

Installation

Option 1: Python (Recommended)

Easiest installation using uvx (no manual install needed):

# uvx automatically downloads and runs the package
loop-mcp

Or install permanently:

pip install loop-mcp

Then add to your MCP configuration:

For Cursor, add to .cursor/mcp.json:

{
  "mcpServers": {
    "loop-engineering": {
      "command": "uvx",
      "args": ["loop-mcp"]
    }
  }
}

For Kiro, add to .kiro/settings/mcp.json:

{
  "mcpServers": {
    "loop-engineering": {
      "command": "uvx",
      "args": ["loop-mcp"]
    }
  }
}

For Claude Desktop, add to ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "loop-engineering": {
      "command": "uvx",
      "args": ["loop-mcp"]
    }
  }
}

Quick Start (3 minutes)

1. Install the MCP Server

uvx loop-mcp

2. Add to Your AI Agent Config

Add the MCP server to your config file (see Installation section above for exact paths).

###3. Restart Your AI Agent

  • Cursor: Reload window or restart app
  • Kiro: MCP servers auto-reconnect on config changes
  • Claude Desktop: Restart the application

4. Start Using It

Simply talk to your AI agent naturally:

You: "Set up a CI triage loop that runs every 6 hours"

AI Agent: [uses create_loop tool]
✅ Loop created: ci-triage
✅ Skill file created at .loop/skills/ci-triage.md
✅ State initialized at .loop/state/ci-triage.json
✅ Scheduled to run every 6 hours

You: "What's the status of my loops?"

AI Agent: [uses list_loops tool]
📊 Active loops:
- ci-triage: Last run 2h ago, 3 PRs opened today, 67% acceptance rate

That's it! The AI agent handles everything using the MCP tools.

Resources

Build the loop. Stay the engineer.