GuideAI agents working together

AI agents working together, in one shared room.

Most tools run AI agents side by side, each in its own window on its own task. Lunchbox puts them in one room on one task. The lead agent splits the work, hands parts of it to the others by name, and they check each other's results until it is done.

Shared agent roomA shared agent room is a desktop app where AI agents from different vendors work in one conversation with one memory, hand tasks to each other, and check each other's results.

AA real room

  1. 00:00You ask for a hackathon launch: find out what people on X want, write the announcement, design the banner and put up the event page.
  2. 00:09Claude makes the plan and asks Grok to search X.
  3. 00:19Grok reports the three things people ask for most, with post counts.
  4. 00:23Claude briefs Codex, and Codex draws the banner.
  5. 00:32Claude writes the announcement from what Grok found.
  6. 00:34Codex builds the event page and runs the tests.
  7. 00:41All 18 tests pass, and Claude says the launch is ready to post.

From an unedited recording of Lunchbox 0.1.1. Watch it

BFour rules that keep it together

The first agent leads

The first agent in your crew answers you. Midway through a task it brings in a crewmate for the part that crewmate does better, with a line like "@codex: build the event page".

Crewmates start briefed

A crewmate gets the goal, what the room has learned, the files that changed and its own task. Only the lead reads the whole log, so nobody pays to read everything twice.

Work moves on at the limit

When an agent runs out of usage, the next one takes the same task with a briefing. Lunchbox shows the reset time and brings the first one back after it.

Claims need proof

An agent can only say the tests pass when a test command passed after its last edit. Without one, the claim waits for you.

CWhat teams of agents are good at

A feature with tests

One agent plans and writes, another builds and runs the tests, and the claim that it works is checked against what ran.

Research, then writing

Grok searches X and the web, Claude turns what it found into the announcement, Codex builds the page.

A second opinion

Ask the whole crew at once and keep the best answer side by side. Your pick feeds that agent's rating.

DQuestions

What is a shared agent room?

A shared agent room is a desktop app where AI agents from different vendors work in one conversation with one memory, hand tasks to each other, and check each other's results.

How do AI agents work together in Lunchbox?

The lead agent splits the task and asks crewmates for parts of it by name. Every agent shares the room's memory, so a crewmate starts with the goal, what is known and its own task, and the agents check each other's results.

Can Claude Code and Codex work on the same task?

Yes. In a Lunchbox room Claude Code can lead and ask Codex to build or test a part, or the other way around, on your own plans.