One window.Every project.
Tabs keeps each project in its own tab, with a code editor, AI agents, Dock, Git and a browser inside.
Scroll to open the window
Step inside.
- 01 Each project is a tab
- 02 Its tools, in the toolbar
- 03 Your choice of agent
Code OSS editorProject conversationsPresets and servicesVisual GitBuilt-in browserAgent CLIs, each in its own tabAny website as a tab
CodeAgentsDockGitBrowserCLIWeb apps
A real Code OSS editor with native file access and VS Code extensions.
Talk to Claude, Codex, Cursor and more about the project that's open.
Your dev servers and CLIs, side by side. Save install, start and preview as a preset and run it from one button.
Review diffs, branches, commits and pull requests without leaving the tab.
Preview localhost with devtools, and let your agents see and click the same page.
Run Claude Code, Codex or any command line in its own tab, right beside your code. Or flip a thread from chat to the agent's CLI with ⌘⇧J.
Pin Figma, Linear, ChatGPT or your staging app to the toolbar. Each one stays signed in, per project.
Sub-agents and Overmind
One stays in line. One branches out.
- Where
- Inside the chat you are already in.
- Good for
- A review, a lookup, a quick fix.
- Comes back as
- An answer in the same conversation.
- Where
- In its own threads, each on its own branch and worktree.
- Good for
- A whole release or a large feature.
- Comes back as
- Branches merged and tested, and one report.
Overmind, step by step
Follow one goal down the page.
goalShip v2: team plans, SSO and a new dashboard
01 · Plan
It starts with a plan.
The lead agent reads your code and turns the goal into eight clear tasks, within the token budget you set.
02 · Split
Every task gets its own thread.
Each thread works on its own git branch and worktree, so nothing collides.
03 · Run
All eight run at the same time.
Each task goes to the model you chose for that kind of work.
Steer
You stay in control.
Open any thread to watch it, send a message to steer it, or stop it. The others carry on.
Retry
If one fails, it tries again.
Billing broke two tests. Overmind retried it on the same branch, and it passed.
Wait
Tasks wait their turn.
End-to-end tests need billing and SSO, so they start only once both are finished.
Budget
It stays within budget.
This run used 9.4M of the 12M tokens you allowed.
04 · Merge
Everything merges back.
All eight branches are merged, then the full test suite runs. 212 tests passed.
Report
You get one report.
What shipped, what changed and what to look at first.
Only in Tabs
The best model for every job.
In Tabs any agent can hand work to an agent from any other provider. So put your favourite in charge, then give each kind of work to the model that does it best.
01Lead
Your favourite. It plans the work and runs the whole operation.
02Design
The one with the best eye for interfaces.
03Research
The one you trust to read everything first.
04Build
The one whose code you'd merge without a fight.
05Review
A second opinion, from a different maker.
06Test
The thorough one. It tries to break it.
7,529,536 possible crews, all on plans you already pay for.
- Claude
- Codex
- Antigravity
- Cursor
- Copilot
- Grok
- OpenCode
- Kilo
- Factory Droid
- OpenRouter
- Kiro
- Muse Code
- Pi
- fx
Fourteen agents · the plans you already pay for
Bring your own models.
- 01 Claude
- 02 Codex
- 03 Antigravity
- 04 Cursor
- 05 Copilot
- 06 Grok
- 07 OpenCode
- 08 Kilo
- 09 Factory Droid
- 10 OpenRouter
- 11 Kiro
- 12 Muse Code
- 13 Pi
- 14 fx
Already pay for Claude Max, ChatGPT, Copilot or an API key? Sign in once and Tabs runs every chat, sub-agent and Overmind task on those plans. No new subscription, no markup.