One workbench for every agent.

Claude Code, Codex, and Antigravity in a single dashboard — live status for every session, smart code reviews, and full control from your phone. Agent Manager takes the juggling out of agentic development.

Every agent, every project, one screen.

Each agent lives in its own terminal window. Run a few of them across a couple of projects and you're playing terminal-tab roulette — which tab was the Codex run in, again?

Agent Manager puts Claude Code, Codex, and Antigravity side by side, grouped by project, with live status for every session. You can see who's working, who's waiting on you, and who's idle — at a glance.

Agent Manager dashboard showing a live agent session with the sessions sidebar grouped by project

Your agents don't stop when you leave your desk.

Step away from your laptop and every agent stalls, waiting on an approval that isn't coming. Agent Manager's full dashboard works from your phone, so the work keeps moving wherever you are.

Review a diff on the couch. Approve a plan from the coffee line. Dictate the next task by voice on a walk. We ship about half of our own code this way.

Agent Manager on a phone: a live agent session with status, review progress, and a command bar

Reviews that turn into fixes.

Agents write code fast — faster than you can read it. Agent Manager runs smart code reviews and streams structured findings in as the review runs: severity, category, impact, and the exact line.

Triage each finding — fix, explain, or close — then send the batch back. Agents pick up the accepted findings and fix them in parallel while you move on.

The smart review panel: a triaged finding with severity and category chips, impact, a suggested fix, and triage progress

From GitHub issue to pull request, hands off.

Point an agent at a GitHub issue and it comes back as a pull request. Agent Manager spins up an isolated git worktree, launches the agent, and tracks the session until the PR is open.

Because every task gets its own worktree, you can run a dozen agents on the same repo in parallel without them stepping on each other — and without losing track of any of them.

The GitHub Issues panel listing open issues for a project, with an agent already working on one

Your workbench, your vibe.

A theme engine ships built in — more than two dozen themes, from Bloomberg-terminal black to QBASIC blue to Windows 3.11 grey, plus a live editor to build your own. Switch anytime; the whole workbench follows.

Agent Manager in the Moon Prism theme — a soft pink pastel interface
moon-prism
Agent Manager in the QBASIC theme — bright blue retro DOS styling
qbasic
Agent Manager in the Future Funk theme — vivid purple vaporwave styling
future-funk
Agent Manager in the Windows 3.11 theme — grey beveled retro desktop styling
win311

Built by people who live in this workflow.

Live session monitoring

Stream agent output in real time. Jump into any session to see what an agent is doing without interrupting its work.

Multi-provider by design

Claude Code, Codex, and Antigravity in one place. Pick the right agent for each task instead of committing to one vendor's workflow.

Voice-first mobile

Dictate the next task, answer an agent's question, or kick off a review by voice — the phone dashboard is a first-class client, not an afterthought.

Parallel worktrees

Spin up isolated git worktrees with one click. Run many agents on the same repo simultaneously without conflicts or lost sessions.

Common questions

What is Agent Manager?

Agent Manager is one workbench for your AI coding agents. It runs Claude Code, Codex, and Antigravity side by side, grouped by project, with live status for every session — and the full dashboard works from your phone.

Which AI coding agents does it support?

Claude Code (Anthropic), Codex (OpenAI), and Antigravity (Google's successor to Gemini CLI). Agent Manager is multi-provider by design, so you can pick the right agent for each task and run them side by side on the same project.

Can I really use it from my phone?

Yes. The full dashboard works on mobile — review a diff, approve a plan, or dictate the next task by voice. We ship about half of our own code this way.

How do code reviews work?

Kick off a review and structured findings stream in as it runs — each with severity, category, and impact. Triage them one by one, then send the batch back and agents fix them in parallel.

Is Agent Manager available now?

Agent Manager is currently in early access. Sign up with your email and we'll reach out with next steps.

Get started with Agent Manager

Enter your email and we'll reach out with next steps.