"Vibe Coding" LangGraph apps with llms.txt and MCP

llms.txt is an emerging standard to provide LLMs with information about a website with brief background information, guidance, and links to detailed markdown files. Here, we create a simple MCP server to expose llms.txt to any MCP host (Claude Code/Desktop, Cursor, Windsurf) and make tool calls to fetch / read URLs from from llms.txt. We introduce llms.txt files for LangGraph and use them with our MCP server to fetch docs as needed by any MCP host code agent, making retrieval visible to the user and easy to audit. Repo: https://github.com/langchain-ai/mcpdoc LangGraph llms.txt files: https://langchain-ai.github.io/langgraph/llms.txt https://langchain-ai.github.io/langgraphjs/llms.txt MCP video: https://www.youtube.com/watch?v=CDjjaTALI68 Video notes: https://mirror-feeling-d80.notion.site/MCP-Server-for-llms-txt-1ba808527b1780b38388ee8126933592?pvs=4 Timestamps: 00:00 - Introduction to LLMs.txt Standard 00:20 - Exploring LangGraph LLMs.txt Documentation 00:33 - Demo Overview: Using LLMs.txt with Applications 00:44 - Demo: LLMs.txt with MCP in Cursor 01:27 - Demo: LLMs.txt with MCP in Windsurf 01:50 - Demo: LLMs.txt with MCP in Claude Desktop 02:23 - Explaining the Problem: IDEs and Document Loading 02:52 - Cursor's Built-in Document Loading Demo 03:56 - Limitations of Built-in Document Tools 04:11 - Approaches for Adding Context to LLMs 04:23 - Context Stuffing: Pros and Cons 05:01 - Vector Indexing: Pros and Cons 05:28 - LLMs.txt as Reference + Tool Calling Approach 06:02 - Using MCP to Connect Tools to Applications 06:47 - Setting Up the MCP Server 07:20 - Testing the MCP Server with Inspector 08:00 - Connecting the MCP Server to Cursor 08:46 - Demo: Using MCP Tools in Cursor (Revisited) 09:13 - Connecting MCP Server to Windsurf 09:26 - Connecting MCP Server to Claude Desktop 09:36 - Demo: Using MCP with Claude Code 10:03 - Conclusion and Benefits of This Approach 10:36 - Final Thoughts and Resources