File Context

You deploy and run a self-contained MCP server that combines AI-powered recruitment capabilities with robust file system tools. This server lets you interact with a Claude-powered recruitment assistant, perform file operations, and securely expose endpoints for integration with external tools, all through a standardized MCP interface.

How to use

You use an MCP client to connect to the server and access four core areas: Claude AI recruitment chat, file operations, candidate evaluation, and project analysis. Open the MCP UI or call the HTTP endpoints to chat with Claude while providing relevant file context, run direct file operations, and submit candidate data for evaluation. You can configure cross-origin permissions so trusted tools can interact with the server securely.

How to install

Prerequisites: ensure you have Node.js 16.0.0 or higher installed on your system.

1. Install dependencies and set up the environment.

2. Configure the Claude API key in a local environment file.

3. Build and run the server in HTTP production mode.

npm install

# If behind a corporate proxy/firewall, configure npm to use a mirror registry
npm config set registry https://registry.npmmirror.com

# Set Claude API key in a local environment file
echo "ANTHROPIC_API_KEY=sk-your-anthropic-api-key" > .env.local

# Build the project
npm run build

# Start the production HTTP server
npm run start:http

# Access the MCP server at http://localhost:3000/mcp and the web UI at http://localhost:3000

Configuration and security notes

Security is handled via CORS and environment isolation. Only trusted origins are allowed to access the API endpoints, and the server is designed to prevent directory traversal and restrict file access to markdown resources.

API endpoints and usage

Chat and AI features are exposed at the chat endpoint, candidate evaluation at the public evaluation endpoint, and file operations at the tools endpoint. Use these endpoints to integrate with external tools and automate workflows.

Testing and development

Run the complete test suite and use the coverage and watch modes during development to ensure stability as you change code or add integrations.

Available tools

read_file

Read a file from the local filesystem. Input is a path and results in the file contents with optional syntax highlighting.

list_directory

List a directory’s contents with metadata such as size and type indicators.

search_files

Perform advanced file pattern matching across a directory tree, returning matching file paths.

analyze_folder

Analyze a folder to produce a structural overview, including file types and project insights.

chat

Claude AI chat interface with file context to discuss code, files, and project structure.

evaluate-candidate

Public endpoint to evaluate a candidate using the Claude-powered recruitment assistant and configured criteria.