study8677/repobrain

🧠 RepoBrain (formerly Antigravity) — Give your repo a brain. ChatGPT for your codebase: works in Claude Code, Cursor, Codex, Windsurf & more.

What it solves

RepoBrain is a repository knowledge engine designed to help developers understand unfamiliar codebases, trace implementations, and perform grounded Q&A. It solves the problem of AI agents forgetting coding styles, guessing at architecture, or hallucinating when they rely on generic search tools like grep or massive, monolithic documentation files.

How it works

RepoBrain creates a persistent knowledge layer stored in a .repobrain/ folder within the project. It uses a multi-agent cluster via rb-refresh to autonomously read the source code and generate specialized knowledge documents for each module. When a user asks a question via rb-ask, a router directs the query to the appropriate ModuleAgent, which provides an answer grounded in actual source code, including specific file paths and line numbers.

Who it’s for

It is intended for developers using AI-powered IDEs (such as Cursor, Windsurf, VS Code + Copilot, or Claude Code) and AI agents (like Aider or Cline) who need a consistent, cross-IDE knowledge base for their codebase.

Highlights

  • Cross-IDE Compatibility: Knowledge is stored in a local folder, making it portable across different IDEs and LLM providers.
  • Grounded Answers: Provides answers with direct references to source files and line numbers to reduce hallucinations.
  • Multi-Agent Architecture: Uses a swarm of agents to decompose the repository into modules and self-learn the architecture.
  • MCP Support: Can be integrated as an MCP (Model Context Protocol) client/tool for other agents.
  • Incremental Updates: rb-refresh updates only affected groups after committed changes, preserving existing knowledge during failures.

Related

  • Project
    sourcebot-dev/sourcebot

    A self-hosted code context layer that indexes multiple repositories for centralized code search and AI agent integration via MCP.

  • Project
    repowise-dev/repowise

    Repowise is a self‑hosted Python tool that indexes a repository’s call‑graph, git history, tests, docs and design decisions into a local database. It powers zero‑LLM, token‑saving queries for developers and coding agents—answering questions about call relationships, change risk, dead code, documentation drift, security secrets, and more. Install with `uv tool install repowise`, run `repowise init` to build the index, and `repowise serve` to expose a dashboard and MCP endpoint. Benchmarks show ~31 % fewer agent tokens and 2.3× more defect detection versus competing tools.

  • Project
    hechtcarmel/jetbrains-index-mcp-plugin

    A JetBrains IDE plugin that runs a local Model Context Protocol (MCP) server, exposing the IDE’s indexing, refactoring, diagnostics, and build capabilities as JSON‑RPC tools for AI coding assistants (Claude, Codex, Cursor, etc.). Supports multiple languages, headless operation, and optional project‑lifecycle management.

  • Project
    NateBJones-Projects/OB1

    An infrastructure layer that provides a single, shared persistent memory database with vector search for all your AI tools, eliminating fragmented data across different AI clients.

  • Project
    activeloopai/hivemind

    Hivemind is a shared‑memory platform for AI coding assistants (Claude Code, OpenClaw, Codex, Cursor, Hermes, pi, etc.). It records every prompt, tool call, and response in a Deeplake database, mines the logs for recurring patterns, creates reusable `SKILL.md` files, and makes those skills instantly available to all agents. Benchmarks show 25 % lower cost, 1.7× fewer tokens, and 31 % fewer dialogue turns versus a no‑memory setup. Installation is a single script or npm package that auto‑injects hooks into supported assistants, and the system can run on any cloud storage you control.