Skip to content

Latest commit

 

History

History
80 lines (68 loc) · 3.3 KB

AGENTS.md

File metadata and controls

80 lines (68 loc) · 3.3 KB

Project Overview

MCP server: mcp-factory-platform-server

MCP Factory Platform Server - Agent interface for platform management

This is an MCP (Model Context Protocol) server built with mcp-factory. It provides tools, resources, and prompts accessible via the MCP protocol.

Server Configuration

Claude Desktop Configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "mcp-factory-platform-server": {
      "command": "uv",
      "args": ["run", "python", "server.py"],
      "cwd": "/path/to/mcp-factory-platform-server"
    }
  }
}

Direct Connection:

# Navigate to project directory first
cd /path/to/mcp-factory-platform-server
uv run python server.py

Note: For alternative configuration methods (different environments, authentication, etc.), see the MCP Configuration Guide or consult your MCP client documentation.

Dev environment tips

  • Navigate to project directory: cd workspace/projects/mcp-factory-platform-server
  • Start server: uv run python server.py
  • Install dependencies: uv sync
  • Format code: uv run ruff format .
  • Check linting: uv run ruff check .
  • Add new tools in tools/ directory with @server.tool() decorator
  • Tools are auto-discovered from tools/, resources/, prompts/ directories
  • Configure environment variables in .env file (copy from env.example)

Testing instructions

  • Use mcp-inspector-server tools for all testing
  • Call inspect_mcp_server to verify server connectivity and registration
  • Use comprehensive_server_test for complete validation of all tools
  • Test individual tools with call_mcp_tool
  • Run uv run ruff check . before committing - all checks must pass
  • Ensure all tools have comprehensive docstrings (required for MCP)
  • Test with production backend: PLATFORM_API_BASE=https://mcp-factory-beta.up.railway.app

PR instructions

  • Title format: [mcp-factory-platform-server] <Description>
  • Always run uv run ruff check . and uv run ruff format . before committing
  • Add or update tests for new/modified tools
  • Update CHANGELOG.md with your changes
  • All tool functions must return dict[str, Any] for MCP compatibility
  • Ensure all API calls are properly error-handled

Code Style Guidelines

  • Use uv run ruff format . to format code (from project directory)
  • Run uv run ruff check . before committing
  • Use type hints: functions return dict[str, Any] for tools
  • Add docstrings to all functions (required for MCP registration)
  • Use direct function parameters, not Pydantic models
  • Follow async/await patterns for all API calls

Component Management

This project supports dynamic component discovery and registration:

Adding Components:

  • Tools: Create .py files in tools/ directory with functions decorated with @server.tool()
  • Resources: Create .py files in resources/ directory with functions decorated with @server.resource()
  • Prompts: Create .py files in prompts/ directory with functions decorated with @server.prompt()

Component Discovery:

  • Components are automatically discovered and registered in config.yaml
  • Use descriptive function names and comprehensive docstrings
  • Ensure all parameters are JSON-serializable for MCP compatibility
  • All tools should use platform_api_adapter.py for backend API calls