Pricing
Plans
Hobby is free. Pro and Team add a server sandbox for tests that need real Node.
The editor and agents, on your own key.
$0 / mo
No card required
- Composer, Chat, and Inline — not gated
- Plan before the first diff
- One send = one call on your key
- Up to five bring-your-own keys
- Tab ghost-text — fast model, on your key
- Tests run free in your browser
- One background job
- ACP: Claude Code, Codex, OpenCode, Grok Build in the same diffs
- Session cap on by default
- Download a zip of the project
Everything in Hobby, and a server sandbox for tests that need real Node.
Not for sale
Not available yet. Nothing is charged.
- Unlimited Composer on your GPT, Claude, Gemini, DeepSeek, Grok keys
- You pick the model. No Auto
- Tests that need real Node run in a sandbox (300 a month)
Headroom for people who live in the agent.
Not for sale
Not available yet. Nothing is charged.
- Everything in Pro
- Tests that need real Node run in a sandbox (1,000 a month)
- Three concurrent background jobs
Paid plans are not available yet: there is no checkout, and no card is charged. Bring your own key
What the plan actually limits
- Composer, Chat, and Inline on Hobby, Pro, and Team.
- One send is one call on the key you attached.
- You pick your Grok, GPT, Claude, Gemini, or DeepSeek key. No silent Auto.
- Or a custom OpenAI-compatible endpoint, such as Ollama, LM Studio, or OpenRouter.
- Session cap on by default: about $1 on your keys per 24 hours. Raise it in Settings.
- Tab ghost-text uses a fast model on your key, not grok-4.5 per keystroke.
- Tests run free in your browser on every plan:
node --test, Vitest, or Jest, for projects without a build step or a DOM. - Tests that need real Node run in your own CI: Agent Check runs them on your GitHub Actions runner. Pro adds a server sandbox for them.
- The public demo plays recorded runs. It calls no model and uses no key.
- Your provider bill is theirs. We do not markup tokens on a key you attached.
Go to model settingsSet up Agent Check