COOKREW an open-source spatial workspace for AI agents

Cookrew / Features / canvas

A spatial canvas for AI agents: terminals, notes and browsers

The Cookrew canvas is an infinite spatial workspace where each agent is a card with a live terminal, a name, a role and a turn tracker; sticky notes hold specs and handovers; browser cards are real Chromium tabs an agent can drive; and wires between cards say who hands work to whom.

Get CookrewGet started

RECORDED3 steps, captured from the running app

How it looks

  1. 1

    The crew at work on one canvas

    The crew at work on one canvas
    ● RECFive agents on three harnesses — Claude Code, Codex and Pi — working a scratch TypeScript repo together: each card carries the brief it was given, what it answered and how many checkpoints it has taken, with the note one of them wrote and the page it opened sitting under its own card.
  2. 2

    A card opened, its rail scrubbed

    A card opened, its rail scrubbed
    ● RECClicked Mason open and dragged its checkpoint rail up from the live end: the rail fanned out into the eight turns it had taken — the test runs, the rename, the README line, the commit — and the transcript on the left followed the drag to the turn under the marker.
  3. 3

    An agent stops to ask, in its own prompt

    An agent stops to ask, in its own prompt
    ● RECQuill reached a decision it would not take alone — what the cut looks like when text is truncated — and raised Claude Code’s own question prompt for it. The card went to NEEDS ATTENTION and the work stopped there, with the two options named and nothing chosen.

Questions about canvas

Is the terminal on a card a real terminal?

Yes. Each card is a PTY hosted by tmux or herdr, running the harness’s own CLI. What you see is what the agent sees.

Can an agent use a browser card?

Yes. Browser cards are Chromium pages driven over the DevTools protocol; an agent can navigate, click, fill and read them with the cookrew browser commands.

Related

Run Claude Code, Codex, OpenCode and Pi in one team

Claude Code, Codex, OpenCode, Pi and a plain shell ship in the dock. Add by link brings somebody else’s crew into your dock.

Every agent turn is a checkpoint: scrub, fork, pin

Scrub back through a session, fork a new agent from any past turn, pin a version when a team is exported or called.

The Board: every agent’s phase, tokens and checkpoint on one screen

What each agent is doing, how many tokens it has spent, which checkpoint it is on. Notes, forks and session cards mix into one activity stream.

Cookrew is an open-source desktop workspace for running a team of AI coding agents on one canvas. It runs Claude Code, Codex, OpenCode and Pi side by side, records every agent turn as a checkpoint, and can serve a saved team at a cookrew.dev address where anyone can open a live session from a browser or from the Cookrew app. Source on GitHub ↗