Features
- 💯 Free and open source.
- 🖥️ Uses your own
claude,codex,opencodeorpilogin. No API key needed. - ☁️ Easy to set up on a VPS, so your agents keep working when your laptop is closed.
- 📱 Fully responsive. Start and review tasks from your phone.
- 🔀 Every task gets its own git worktree, so several agents can work at the same time. Extra tasks wait in a queue.
- 🤖 Turn on Autonomous and a run never stops to ask. It just finishes.
- 📡 Watch it work live: agent text, tool calls, tokens and cost.
- 🏁 Run the same task ×2 or ×3, compare the diffs and keep the best one.
- 🧩 Skills are Markdown files and workflows are short YAML files. Mix agents per step.
- 🐙 Run the agent straight on a GitHub issue. Nothing merges on its own.
- 📂 One cockpit for all your projects.
- 💾 No database. Everything is saved as plain files in
.ai/cezar/.
Screenshots
Parallel tasks — Run and queue many tasks, each in its own git worktree.
Live run — Every step, tool call and token, as it happens.
Variants — Run a task ×2 or ×3 and keep the best diff.
Workflows — Drag skills and checks into a chain, saved as YAML.
GitHub — Hand an open issue to the agent in one click.
Skills + Autonomous — Pick a playbook, flip Autonomous and walk away.
On your phone — the same cockpit, from the task list to the diff.
Quick start
You need Node 20+ and at least one agent CLI you're logged into:
Claude Code, Codex,
OpenCode or pi.
git and gh are optional.
cd your-repo
npx cezar-cli
This opens the cockpit at http://localhost:4321. Type a task, pick a workflow, then hit Start.
npx cezar-cli run "add a --json flag to the export command" # headless, no browser
npx cezar-cli init # scaffold .ai/cezar/
npx cezar-cli@nightly # try tonight's build
Just want to look around? Run
CEZ_DRY_RUN=1 npx cezar-cli. It uses a built-in mock agent, so you don't need to log in.
Run it on a server
npx cezar-cli server-install --platform ubuntu-vps
This sets up HTTPS, a login and a system service, so you can open the cockpit from anywhere, including your phone. There are guides for Ubuntu VPS and macOS + ngrok.
How it works
- You describe a task. Type it, attach files, or start from a GitHub issue.
- cezar runs a workflow (agent steps plus shell checks) in a new git worktree, using your agent CLI.
- The cockpit streams every step live. If a check fails, the agent tries again and sees the error.
- You check the result. Read the diff, send notes back, or open a draft PR.
A workflow is a small YAML file in .ai/cezar/workflows/:
name: fix-and-verify
steps:
- id: implement
prompt: "{{task}}"
skill: project-conventions # optional: a Markdown skill from .ai/skills
runner: codex # optional: which agent runs this step
- id: verify
command: "npm test" # exit 0 = pass
onFail: { retry: implement, max: 2 }
The built-in quick-task workflow runs with no setup.
Documentation
The reference covers everything else: configuration, environment variables, agent backends, multiple projects, remote access and local development.
Contributing
- Found a bug or missing something? Open an issue.
- Want to contribute? PRs are welcome. See local development to get started:
git clone https://github.com/open-mercato/cezar.git && cd cezar
npm install
npm run dev
License
MIT © Patryk Lewczuk. Full text in LICENSE.