RepoPilot

Introduction

RepoPilot indexes Git repositories and turns them into engineering intelligence — real dependency graphs, search, impact analysis, and evidence-backed answers.

What RepoPilot does

GitDiagram shows a sketch. RepoPilot indexes your repo and keeps engineering intelligence behind every view.

Paste any public GitHub URL on the home page, or connect GitHub to index private repositories. RepoPilot parses source files with Tree-sitter, builds module and symbol dependency graphs, indexes searchable code chunks with embeddings, and ingests git history for hotspot analysis.

Core capabilities

  • Real dependency graph — Built from import edges and AST analysis — not AI guesses.
  • Impact analysis — See what breaks before you merge a change.
  • Code hotspots — Find where the team actually churns code.
  • Ask with citations — Questions answered with file-and-line evidence.
  • PR review findings — Evidence-backed review signals, not vibes.
  • MCP for agents — Connect IDE agents to search, impact, Ask, and context tools.

How it is organized

RepoPilot is a Yarn monorepo with three packages:

  • @repopilot/web — Next.js dashboard and public marketing shell (port 3000)
  • @repopilot/api — Fastify REST API, indexing pipeline, and MCP server (port 3001)
  • @repopilot/common — shared types and GitHub URL helpers

PostgreSQL (with pgvector) stores indexed data. Redis queues background index jobs. A separate worker process drains the queue in production; local dev can run indexing inline with INDEX_INLINE=true.

Where to go next