Reposwarm

How do you give AI coding agents context across multiple repositories?

RepoSwarm analyzes repositories and generates standardized .arch.md architecture files.
It commits the results to a shared hub repository that coding agents can read as context.

Add that hub to your agent's workspace so it can consult architecture documentation across repositories.
Architects and developers can use the same output for onboarding and architecture reviews.
Generated findings need human review.

- Git providers: GitHub, GitLab, AWS CodeCommit, Azure DevOps and Bitbucket.
- Model backends: Anthropic API, Amazon Bedrock or a LiteLLM proxy.
- Updates: checks repository commits, branches and prompt versions to decide when to re-analyze.
- Caching: reuse prompt results stored in DynamoDB or local files; repository-level update checks use DynamoDB.
- Runs with Temporal workflows. Local setup requires running Docker, Docker Compose and Git.
- License: Apache-2.0.

See generated output · Install and run · How agent context works

  AI-powered multi-repo architecture discovery platform

  Quick Start •
  What Is It •
  How It Works •
  Ecosystem •
  Contributing
📦 Previously loki-bedlam/repo-swarm. Moved to the RepoSwarm organization. Old URLs redirect automatically.

Quick Start

Install the CLI and bootstrap everything in one shot:

Then:

That's it. The CLI handles setup, configuration, investigation, diagnostics, and results — all from a single binary.

Prerequisites

- Docker must be running (not just installed) — all services run as containers
- Git — for cloning repositories during investigation

Bedrock Users

If using Amazon Bedrock, the worker needs these env vars (set automatically by the CLI):
- CLAUDE_CODE_USE_BEDROCK=1
- AWS_REGION=us-east-1 (or your region)
-…