mcp-docs-server
A lightweight MCP server for local documentation - simple alternative to RAG pipelines
Documentation
MCP Docs Server
A lightweight Model Context Protocol (MCP) server that provides direct access to local documentation files - a simple alternative to complex RAG pipelines for project-specific context.
Overview
This MCP server reads a single markdown file (`context.md`) and exposes its contents through two simple tools:
- `get_context_overview()`: Lists all section titles
- `search_context(query)`: Searches content across all sections
Perfect for giving LLMs access to project documentation without the overhead of vector databases or embedding models.
Features
- Zero dependencies beyond the MCP Python SDK
- Lightning fast - direct file access, no vector search
- Simple setup - works with both GUI and CLI MCP clients
- Cross-platform - includes shell wrapper for macOS/Linux compatibility
- Robust error handling - comprehensive logging and debugging support
Installation
Prerequisites
- Python 3.11+
- MCP client (Claude Desktop, cline, etc.)
Setup
1. Clone this repository:
git clone https://github.com/unlock-mcp/mcp-docs-server.git
cd mcp-docs-server2. Install dependencies:
pip install -r requirements.txt3. Create your documentation file:
# Create a context.md file in the project root with your documentation
echo "# My Project Docs\n\nThis is my documentation." > context.mdClient Configuration
For Claude Desktop (GUI clients)
Use the shell wrapper for reliable execution:
1. Make the wrapper executable:
chmod +x run_context_server.sh2. Install the server:
mcp install ./run_context_server.sh --name "docs-server"For cline and CLI clients
Add to your MCP configuration file:
{
"mcpServers": {
"docs-server": {
"timeout": 60,
"type": "stdio",
"command": "/opt/homebrew/bin/python3.11",
"args": [
"/path/to/mcp-docs-server/mcp_context_server.py"
],
"env": {}
}
}
}*Note: Update the Python path to match your system (`which python3.11`)*
Development
Testing
Use the MCP development tools for easy testing:
mcp dev ./run_context_server.shThis launches a web-based inspector for testing your server.
Debugging
The server logs to stderr for debugging. Check your MCP client's logs if you encounter issues.
Common issues:
- ENOENT errors: Use the shell wrapper or specify full Python path
- Import errors: Ensure `mcp[cli]>=1.2.0` is installed
- File not found: Verify `context.md` exists in the project root
File Structure
mcp-docs-server/
├── mcp_context_server.py # Main server implementation
├── run_context_server.sh # Shell wrapper for GUI clients
├── requirements.txt # Python dependencies
├── context.md # Your documentation (create this)
└── README.md # This fileUsage
Once configured, you can use these tools in your MCP client:
- Get overview: "What sections are available in the docs?"
- Search content: "Search for authentication in the docs"
- Specific queries: "How do I configure the database?"
Tutorial
For a complete walkthrough of building this server from scratch, including common pitfalls and solutions, see the full tutorial: Ditching RAG: Building a Local MCP Server for Your Docs
Contributing
Contributions welcome! Please feel free to submit issues and pull requests.
License
MIT License - see LICENSE file for details.
Built with ❤️ by the UnlockMCP team.
Frequently asked questions
What is mcp-docs-server?
mcp-docs-server is A lightweight MCP server for local documentation - simple alternative to RAG pipelines
How do I install mcp-docs-server?
Open the GitHub repository and follow its README. Most MCP servers are added to your client's MCP config, then called by your agent.
Is mcp-docs-server open source?
Yes — it is hosted on GitHub at https://github.com/Unlock-MCP/mcp-docs-server.
Related MCP tools
Cognee is the open-source AI memory platform for agents. Give your AI agents persistent long-term memory across sessions with a self-hosted knowledge graph engine.
Automate browser based workflows with AI
Hindsight: Agent Memory That Learns
A privacy-first app that strips AI watermarks from content you own.
Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc.
The power of Claude Code / GeminiCLI / CodexCLI + [Gemini / OpenAI / OpenRouter / Azure / Grok / Ollama / Custom Model / All Of The Above] working as one.
Run your own MCP server? See who uses it and what to fix.
Measure it with TrackMCP