PT-MCP

PT-MCP helps AI coding assistants understand your codebase by providing rich context through integrated knowledge graphs, semantic schemas, and detailed codebase analysis. It answers the essential question of where you are now in your project, so your AI tools can operate with accurate, up-to-date context and generate meaningful code recommendations.

How to use

You use PT-MCP by running a local MCP server that analyzes your project and exposes context data to your AI tooling. The server can generate context files in multiple formats, track changes incrementally, and expose an API surface for your integration. To start using it, configure your MCP client to connect to the context manager and request the kinds of context data you need, such as codebase structure, dependencies, and patterns. You can then reuse this context to guide code completion, refactoring, and architectural decisions.

How to install

# Prerequisites
- Node.js 14+ installed
- npm available in your environment

# Install dependencies
npm install

# Build the MCP server
npm run build

Configuration and usage notes

The server is designed to be consumed by an MCP client. Here is a typical local configuration example that connects a client to the context manager using a local stdio server setup. This config tells the client how to launch the MCP server and where to find its main entry point.

{
  "mcpServers": {
    "context_manager": {
      "command": "node",
      "args": ["/path/to/context-manager-mcp/dist/index.js"],
      "env": {}
    }
  }
}

Additional sections

Security: Keep the server and client communications within trusted networks. Use environment-specific configurations to limit exposure of analysis data. If you enable external access, implement proper authentication and access controls.

Notes: The MCP server currently provides core analysis tools with stub implementations for some features. You can rely on the foundational tools for codebase analysis, context file generation (in future), and incremental updates as development progresses.

Troubleshooting: If context data appears stale after changes, trigger a full regeneration or run incremental updates when files change. Check that the MCP process has the necessary permissions to read your project files and write context outputs.

Available tools

analyze_codebase

Perform comprehensive codebase analysis including structure, dependencies, and metrics.

generate_context

Generate context files in specified format for downstream clients.

update_context

Incrementally update existing context files based on code changes.

extract_patterns

Identify and extract architectural and coding patterns.

analyze_dependencies

Analyze and map internal and external dependencies.

watch_project

Start monitoring project for changes and auto-update context.

extract_api_surface

Extract and document public API surface.

validate_context

Validate accuracy and completeness of generated context files.