It plans out loud. You approve once, then it works.
Every agent starts in plan mode: it reads, writes a plan, and waits. Approve it and the agent runs the whole plan — it works in a container of its own, so it never stops to ask the same yes twice.
- 2 answersapprove the plan · or reply to keep planning
- Per turnswitch model, harness and reasoning effort mid-conversation
- Your accountit runs on your Claude, ChatGPT or xAI subscription
The plan is the contract
Before it edits anything, the agent shows which files it will touch and what it will do to each. Approve it and the plan becomes a to-do list you watch.
- Thinking, tool calls and to-dos stream as they happen, never summarised after the fact
- Steer mid-turn: a message lands in the running turn instead of queuing behind it
Permission is a dial, not a switch
The same conversation can be strict on a migration and loose on a changelog — the composer's mode pill sets how much the agent may do unattended, per turn. An approved plan is the exception: you have read what it intends, so it runs the lot.
Bring your own model
Claude Code, Codex, Grok, Kimi Code and Gemini, on subscriptions you already pay for. Connect once; the credential stays in the sandbox and every turn runs on your plan.
The same conversation on your phone
The mobile shell is the same app, not a companion. An agent that stops for a decision at 11pm is one tap from unblocked.
