async-code

Why we included this project

Waiting on one Claude Code session to finish before you can start the next one is a familiar bottleneck, and this project attacks it directly by letting you run several agents at once. It puts Claude Code and other CLI agents behind a Codex-style web interface where you can queue up tasks, watch them run in parallel, and compare what different models produce before you pick a winner. Each task gets its own container, so the work stays sandboxed, and the whole thing deploys with docker-compose if you want to host it yourself. Git integration handles cloning, committing, and opening pull requests, which means a task can go from prompt to PR without you hovering over the terminal. For teams that want a self-managed alternative to hosted agent platforms, or that need to run the same workload against several agents to see which one performs best, the comparison workflow is the part that earns its keep.