Customized Model Context Protocol Document Interface

An enhanced implementation of the MCP server designed to bridge LLM execution environments (such as Cursor, Claude Desktop, Windsurf) with designated knowledge sources, operating under the Model Context Protocol specification.

Core Functionality

This server furnishes MCP-compliant host platforms with the following capabilities: 1. Retrieval access to curated documentation assets (specifically langgraph.txt and mcp.txt). 2. Utility functions designed to ingest content from embedded URLs within those foundational documents.

Accessible Knowledge Sets

Currently configured to serve: - Documentation pertaining to LangGraph (sourced from: https://raw.githubusercontent.com/esakrissa/mcp-doc/main/docs/langgraph.txt) - Documentation detailing the Model Context Protocol itself (sourced from: https://raw.githubusercontent.com/esakrissa/mcp-doc/main/docs/mcp.txt)

Deployment Guide

Initialization and Execution

bash

Obtain the source repository

git clone https://github.com/esakrissa/mcp-doc.git cd mcp-doc

Set up and activate a dedicated Python environment

python -m venv venv source venv/bin/activate # Use venv\Scripts\activate on Windows systems

Install the package for local development

pip install -e .

Operating the Service

Launch the server utilizing the installed package entry point:

bash

Invoke the server using the configuration file, SSE transport, and specified host/port

mcpdoc \ --json config.json \ --transport sse \ --port 8082 \ --host localhost

Alternatively, leverage the uv package manager for execution:

bash

Install uv if necessary

curl -LsSf https://astral.sh/uv/install.sh | sh

Execute the server via uvx shim

uvx --from mcpdoc mcpdoc \ --json config.json \ --transport sse \ --port 8082 \ --host localhost

IDE Integration Protocols

Integration with Cursor IDE

Configure the following structure in ~/.cursor/mcp.json:

{ "mcpServers": { "mcp-doc": { "command": "uvx", "args": [ "--from", "mcpdoc", "mcpdoc", "--urls", "LangGraph:https://raw.githubusercontent.com/esakrissa/mcp-doc/main/docs/langgraph.txt", "ModelContextProtocol:https://raw.githubusercontent.com/esakrissa/mcp-doc/main/docs/mcp.txt", "--allowed-domains", "*", "--transport", "stdio" ] } } }

Subsequently, integrate these directives into Cursor's Custom Instructions:

When addressing ANY query related to LangGraph or the Model Context Protocol (MCP), invoke the mcp-doc service to augment the response -- + First, execute list_doc_sources to discover accessible artifacts + Next, employ the fetch_docs utility to read the content of langgraph.txt or mcp.txt + Analyze the retrieved URLs contained within langgraph.txt or mcp.txt + Synthesize findings based on the input query context + Execute fetch_docs for any URLs identified as pertinent to the query + Formulate the final answer leveraging this external documentation context

Verification: Test the setup by posing a question about LangGraph or MCP to Cursor, confirming that it utilizes the documentation server's tools for data retrieval.

Security Considerations

Domain access is strictly regulated by default: - External documentation pointers: Only the specifically listed originating domain is permitted by default. - Internal documentation pointers: No external domains are permitted without explicit allowance. - Use the --allowed-domains flag to designate permissible hosts, or --allowed-domains '*' for universal access (exercise caution with this setting).

Source Attribution

This project represents a derivative work, stemming from the original mcpdoc developed by LangChain AI, tailored specifically to provide focused retrieval capabilities for LangGraph and MCP documentation.