termic.dev
Documentation menu

AI code review

Have one agent review another's work: Termic's review dialog spawns a fresh agent, hands it the diff and a review prompt, and streams the review right in the task.


A second pair of eyes catches things the author misses, and that’s as true for agents as for people. Termic’s code review lets you point a fresh agent at a task’s changes and get a streamed review, without leaving the app or assembling the diff yourself.

How it works

Open the Review dialog and pick which agent you want to do the review. Termic spawns a fresh terminal tab in the task, waits for the CLI to be ready, and sends it a review prompt. The agent then fetches the diff itself and starts streaming its review.

The review prompt has a built-in git diff fallback, so the agent retrieves its own diff rather than Termic pre-injecting a stale snapshot. You always get a review of the current state of the working tree.

Mixing agents

Because the reviewer is just another agent, you can deliberately review with a different CLI than the one that wrote the code. Have claude implement and codex review, or vice versa. Different models notice different things, and a cross-model review is a cheap way to widen the net. The reviewer can be any agent in your registry, built-in or one you added.

Where it fits

Code review pairs with the rest of the review surface:

  • Read the change yourself in the diff viewer first, or alongside.
  • Use Broadcast if you want several agents’ takes on the same change at once, rather than one focused review.
  • For worktree work, once you’re happy, Send to main lands it in your checkout.

Last reviewed: June 5, 2026