Each project carries reusable context. Each task is its own agent session — Claude Code or Codex — in its own git worktree. Drive ten at once, see exactly which one needs you, review every diff before it merges. Runs on your Max/Pro login — no API key, no per-token billing.
- Your Claude plan can run more than one session at a time — stop working it one terminal at a time.
- One screen for every project and every task. No tab-juggling.
- Project context is written once and injected into every task. Stop re-explaining your stack.
- A cross-project "N need you" signal shows exactly which session is waiting on you.
- Parallel sessions — every task is an isolated git worktree with its own agent session.
- Diff review → one-click merge — or AI conflict resolution, branch sync, and GitHub PR creation.
- Pick your agent per task — Claude Code or Codex (beta), both on subscription logins.
- Write-once project context — auto-injected into every task; Refresh with AI redrafts it from the repo.
- Session lineage —
/clearhands a summary to a fresh context window; the task lives on. - Reconnect-safe turns — turns run server-side; reload or sleep the laptop and the transcript catches up. Queue follow-ups mid-turn.
- Integrated terminal + dev servers — a real shell per project, plus supervised dev processes with live logs.
- Cost tracking + insights — live per-task spend and a local analytics dashboard.
- Plus: agent-suggested tasks, task dependencies, image attachments, clone from GitHub, recaps, a first-run tutorial.
Watch the session stream — tool calls, edits, questions:
Review the diff next to the chat, then merge (or open a PR):
A real terminal, right in the workspace:
| Agent | Status |
|---|---|
| Claude Code | Fully supported — the reference driver; every feature lands here first. |
| OpenAI Codex | Beta — works end-to-end; we're still working out the kinks. Issues welcome. |
Want another agent? The driver seam is small — see adding a new agent.
Open Insights from the top bar for a local analytics dashboard of what your agents cost and ship: per-day spend and token usage (including cache reads/writes), tasks shipped, and lines merged to base — sliceable by project and agent across 7/30/90-day ranges, with deltas against the prior period. Everything is computed from the local SQLite database in a single fetch, filter changes recompute instantly in the browser, and nothing is sent anywhere.
npm install
npm run dev
# open http://localhost:3000You need Node 18.18+, macOS or Linux, and the claude CLI
(npm i -g @anthropic-ai/claude-code) with a Pro/Max plan. First run opens a setup
wizard that signs Claude in from the browser, then drops you into a 2-minute hands-on
tutorial. Keep ANTHROPIC_API_KEY unset so it uses your subscription, not the API.
Every setting is an env var with a sane default — see .env.example.
One hardened Docker container, built to sit behind an authenticated tunnel:
docker build -t agent-orchestrator .
ORCH_USER=alice ORCH_PORT=10001 docker compose -p orch-alice up -dThe port binds to loopback only — the app hands out a full shell, so put auth in front and never expose it raw. Tunnels, Cloudflare Access, idle sleep, and every config knob: docs/SELF_HOSTING.md.
Don't want to run a server? getoperator.dev is your own always-on instance — works from your phone, zero setup. Same codebase plus a closed-source control plane.
- Architecture — how it works; the agent-driver seam; adding a new agent
- Self-hosting — Docker, auth, configuration, caveats
- Contributing · Security