RepoSwarm iconRepoSwarm

open-source Free Star0k

AI-powered multi-repo architecture discovery platform that analyzes entire codebase portfolios and generates standardized .arch.md documentation for coding agents and developers

agents python self hosted
249 GitHub Stars
5+ Git Providers
Apache 2.0 License

Overview

RepoSwarm automatically analyzes your entire codebase portfolio and generates standardized architecture documentation. Point it at your GitHub, GitLab, CodeCommit, Azure DevOps, or Bitbucket repos and get back clean, structured .arch.md files — perfect as AI agent context, onboarding docs, or architecture reviews. It uses Claude under the hood, runs as a set of Docker containers orchestrated by Temporal workflows, and only re-analyzes repos with new commits via smart caching.

The Verdict

Who Should Use RepoSwarm?

Best For

  • Teams with 10+ repositories needing consistent architecture docs
  • AI agent developers who need structured codebase context
  • Engineering leaders doing portfolio-level architecture reviews
  • Onboarding engineers to large, unfamiliar codebases
  • AWS/Bedrock shops wanting self-hosted AI analysis

Not Ideal For

  • Single-repo projects (overhead not worth it)
  • Teams without Docker available
  • Real-time code search (use Sourcegraph instead)
  • Non-technical stakeholders expecting a SaaS UI

What's Great

  • Generates consistent .arch.md files across every repo in your portfolio
  • Incremental — only re-analyzes repos with new commits
  • Type-aware prompts (backend, frontend, mobile, infra, libraries)
  • Multi-provider: Anthropic API, Amazon Bedrock, or LiteLLM
  • Multi-git: GitHub, GitLab, CodeCommit, Azure DevOps, Bitbucket
  • Parallel investigation across repos
  • Self-hostable, Apache 2.0 licensed
GitHub README

Watch Out For

  • Requires Docker to run all services
  • Self-hosted only — no managed cloud offering
  • Early-stage, hackathon-origin project with limited community adoption so far
  • Claude API costs accrue per repo analyzed
  • DynamoDB Local setup has had auth quirks on fresh installs
GitHub README — Recent Fixes

Pricing

Open Source
Free
Self-hosted, Apache 2.0. Pay only for LLM API usage.
LLM Costs
Usage-based
Anthropic API, Amazon Bedrock, or LiteLLM — your provider, your costs
View all features & details

Key Features

  • AI-powered codebase analysis via Claude
  • Standardized .arch.md output format
  • Incremental re-analysis (commit-based)
  • DynamoDB or file-based caching
  • Temporal workflow orchestration
  • Parallel repo investigation
  • Results search across all architecture docs
  • Repo diff comparisons
  • Results export to local docs

Ecosystem Components

  • reposwarm-cli — setup, investigate, diagnose
  • reposwarm-api — REST API server
  • reposwarm-ui — Next.js dashboard
  • reposwarm-askbox — AI agent for querying docs
  • Core engine — Temporal workflows + analysis

Supported Git Providers

  • GitHub
  • GitLab
  • Amazon CodeCommit
  • Azure DevOps
  • Bitbucket

Analysis Types

  • Backend — APIs, databases, services
  • Frontend — components, routing, state
  • Mobile — UI, device features, offline
  • Libraries — API surface, internals
  • Infrastructure — resources, deployments
  • Shared — security, auth, monitoring

How It Compares

Feature RepoSwarm Sourcegraph Serena MCP
Focus Architecture docs generation Code search & navigation Single-repo agent context
Multi-repo Yes, portfolio-wide Yes No
Output format .arch.md files Code index In-memory context
Self-hosted Yes (Docker) Yes / Cloud Yes
LLM provider Anthropic, Bedrock, LiteLLM Anthropic, others Any
Pricing Free (OSS) Enterprise Free (OSS)
Maturity Early-stage Mature Early-stage

User Reviews

Loading reviews...