Ask any question about any GitHub repository and get precise,
cited answers — powered by a local RAG pipeline you control.
RepoMind goes beyond simple file references and raw vector similarity. We parse your entire codebase to extract semantic groups—like authentication, payments, and data access—so you can trace real architectural purpose.
Chunks and similarity scores. No sense of what connects to what, leaving you guessing about true architectural impact.
Grouped by what the code actually does — traceable through real call and import relationships, directly mapped to your logic.
Everything you need to navigate, analyze, and query complex codebases faster and with total confidence.
Explore an interactive map of files, functions, and classes built from real AST parsing with tree-sitter—not just shallow text similarity.
Combines fast vector search with structural graph traversal. Answers about multi-file logic and deep dependencies never get missed.
Functions and files are automatically grouped by semantic purpose, mapping raw syntax directly to concepts like authentication and payments.
Filter the noise. Instantly toggle between Frontend, Backend, Testing, or Deployment layers to focus entirely on the architectural slice you need.
Use RepoMind directly inside Claude Code, Cursor, and other tools via our built-in MCP server. No separate app or context switching required.
Stop exploring blind. Step sequentially through tagged logic groups to understand how a specific feature was implemented from start to finish.
A fully local MCP server — no cloud, no API keys required.
# 1. Clone the repo and install dependenciesgit clone https://github.com/srishti-rawat/RepoMind.gitcd RepoMind/repomind-mcp && npm install# 2. Build the MCP servernpm run build# 3. Add to your IDE's MCP config (e.g. ~/.cursor/mcp.json)node dist/index.js
Works with Claude Code, Cursor, Windsurf, and other MCP-compatible tools.