Skip to content
ReArch DOCS

Overview

ReArch is a self-hosted platform that gives your team access to AI coding agents running in isolated Docker containers. Each agent works against a real clone of your repository — reading, modifying, and running code in a sandboxed environment.

Key Features

Section titled “Key Features”
  • Isolated environments — Each conversation runs in its own Docker container with VS Code, an AI agent, and your application runtime.
  • Repository integration — Connect GitHub or Bitbucket. No migration needed, no lock-in.
  • MCP tool support — Extend agent capabilities with external tools (GitHub, Sentry, Slack, Stripe, and more) via the Model Context Protocol.
  • Visual verification — Agents can launch your application, take screenshots, and verify UI changes.
  • Multiple LLM providers — Use Anthropic Claude, Google Gemini, OpenAI GPT, or other providers. Switch models per conversation.
  • Cost monitoring — Per-user and per-conversation cost tracking with analytics dashboards.
  • Self-hosted — Runs on your infrastructure. Your code stays in your network.

How it works

Section titled “How it works”
  1. Connect your repositories — Link your GitHub or Bitbucket workspace. ReArch imports repository metadata and builds Docker images containing your code and dependencies.
  2. Describe what you need — Use natural language to tell the agent what to build, fix, or improve. Anyone on your team can do this.
  3. ReArch works in the background — A sandboxed container spins up, the agent writes code, runs tests, and takes screenshots for verification.
  4. Review and ship — Review the diff, commit, push to a branch, and create a pull request — all from within ReArch.

Use cases

Section titled “Use cases”
  • Non-technical teammates shipping features — Product managers, designers, and ops teams describe what they need and receive working pull requests.
  • Bug fixes and maintenance — Point the agent at an issue and let it investigate, fix, and verify the solution.
  • Scaffolding new projects — Spin up pages, components, API routes, and tests with a single prompt.
  • Code reviews and refactoring — Let the agent analyze your codebase and propose improvements, migrations, or cleanups.
  • Prototyping and experimentation — Test ideas without pulling engineers off their current sprint.

Next steps

Section titled “Next steps”