mcp-file-context

File Context Server provides LLM-accessible file system context with capabilities to read, list, search, and analyze code files. It exposes tools such as list_context_files to enumerate directory contents with metadata, read_context to fetch file contents with encoding options, search_context to perform pattern searches with contextual lines, analyze_code to compute quality metrics like complexity and dependencies, and cache_stats to inspect caching performance. Start it via npx file-context-server and invoke the available tools through the MCP interface to retrieve structured results that LLMs can reason over. The server includes built-in real-time caching and supports file-type filtering, regex searches, and configurable context lines to tailor returned snippets for your prompts.

Prerequisites:

• Node.js (recommended LTS) and npm installed on your system
• Familiarity with MCP tooling and the command interface

Installation steps:

1. Install the package from npm:

npm install @modelcontextprotocol/file-context-server

2. (Optional) Install via Smithery for Claude Desktop integration:

npx -y @smithery/cli install @bsmi021/mcp-file-context-server --client claude

3. Verify installation by starting the server (example):

npx file-context-server

4. If you are integrating with an app using MCP, reference the mcp_config accordingly and manage environment variables as needed.

Tips and notes:

• Environment variables:
  • MAX_CACHE_SIZE: cap on cached entries (default 1000)
  • CACHE_TTL: time-to-live for cache entries in milliseconds (default 3600000)
  • MAX_FILE_SIZE: maximum file size in bytes allowed for reading
• Cross-platform path handling has been updated to improve compatibility; end users do not need to take action, but developers should follow POSIX-style globs and normalized paths when extending functionality.
• Tools return structured JSON objects for easy parsing by LLMs. Use cache_stats to monitor performance and adjust cache-related env vars as needed.
• If you encounter FILE_NOT_FOUND or PERMISSION_DENIED, verify the path and file permissions. Ensure the server has access to the requested directories.
• This server is designed to work with multiple file types and supports recursive traversal, context-aware results, and exclusion patterns via the search and read APIs.