Hello, world — and hello from ShipIt
ShipIt is a self-hosted, chat-driven IDE for coding agents. It doesn’t write code itself — it runs Claude Code or Codex through your real engineering loop: every session gets its own isolated container and a live instance of your app, so the agent can build, run, and fix before you preview and merge through PRs, CI, and deploys on your own Git.
Why it exists
Coding agents are the easy part — you already have Claude Code or Codex. The hard part is everything around them. A few things shape how ShipIt feels day-to-day:
- Real app feedback. Every session runs your full app stack in its own container, so the agent tests against the running app instead of guessing.
- Parallel and isolated. Each agent turn runs in its own Docker container. Concurrent sessions can’t step on each other’s files, processes, or installed dependencies, so you can fan work out without cross-contamination.
- Self-hostable on a remote server. ShipIt is Docker-based end to end. Run it on a VPS and your laptop doesn’t need to be open for the agent to keep working.
- First-class previews from Docker Compose. Declare your dev server (and anything else — databases, queues, log tailers) in
docker-compose.yml. ShipIt surfaces each service as an automatic or manual preview inside the app.
One IDE, everything inline
ShipIt collapses the build-plan-review-ship loop into a single chat-shaped surface. PRs, CI status, deploy status, diffs, commits, Linear and GitHub issues, conversation history, terminal, and live preview all render inline — no jumping out to GitHub, your hosting dashboard, or a separate terminal.
What’s next
We’ll use this blog for release notes, deep-dives on the architecture (container per session, isolated clones, Compose-as-preview), and the occasional opinion on AI-assisted development. Subscribe via RSS if you want updates as they land.
Welcome aboard.