redsquirrel5555/the-boardroom — explained in plain English
Analysis updated 2026-05-18
Have separate AI coding agents working on different parts of a codebase challenge and verify each other's claims.
Set up a shared chat room where a human can mention specific agents to wake them and ask questions.
Add a small chairman model that enforces procedure like requiring evidence for claims and writing verbatim minutes.
Keep long-running agent sessions reachable using a listener script that re-arms itself after each wake-up.
| redsquirrel5555/the-boardroom | 00kaku/gallery-slider-block | 3rd-eden/ircb.io | |
|---|---|---|---|
| Stars | 0 | — | — |
| Language | JavaScript | JavaScript | JavaScript |
| Last pushed | — | 2021-05-19 | 2016-11-16 |
| Maintenance | — | Dormant | Dormant |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 2/5 | 2/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires running a separate terminal session with its own coding agent for each board member.
The Boardroom is a small JavaScript project, about 300 lines of code, that lets several AI coding agents and one human argue with each other in real time. Each agent lives in its own terminal, in its own repository, working on its own part of a system, and they can @mention each other in a shared chat room to challenge claims and check each other's work. The idea behind it is that a single coding agent working alone has no adversary, so it can confidently state things that are almost true with nothing to catch the mistake. The Boardroom gives it one: other long running agent sessions that already have hours of real context about their own part of the codebase, who can read a claim and go verify it directly. A board member is not a fresh agent spun up to answer a question, it is an existing terminal session that pauses what it was doing, responds, and goes back to work, which is why its answers carry more weight than a brand new session would. The clever trick that makes this possible is that most agent command line tools re-invoke the agent once a background task it started finishes. So each agent runs a small listener script that waits quietly until someone mentions it by name, then prints the message and exits. That exit acts like a doorbell, waking the agent so it can read the message and respond, after which it must restart the listener or it becomes unreachable. To run it, you clone the repository, install dependencies with npm in the src folder, which only needs express and ws, and start the server. You then define your board members in a JSON file with a name, color, role, and working directory for each one, paste a starter brief into each agent's terminal, and arm each one as a background listening task. A browser page at localhost shows the room where the human types problems and mentions members to wake them. The README also describes typical roles that emerge in a working board: the human who decides and breaks ties, domain owners who live in the room permanently, specialists brought in for specific topics, and an optional chairman, a small cheap model that manages procedure like opening sessions and writing verbatim minutes without deciding anything itself.
The Boardroom lets multiple long-running AI coding agents in separate terminals argue and verify each other's claims in a shared chat room, alongside a human.
Mainly JavaScript. The stack also includes JavaScript, Node.js, Express.
No license information is stated in the README.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.