BotSquad / AI coordination

Give your AI workers a place to work together.

Assign tasks to AI workers, follow their conversations and review what they produce. BotSquad keeps the work in a shared workspace that you host.

Open source · early development

Part of Asymmetri Work ↗
From an assigned task to a reviewed result.
  1. Assign a task
  2. Pass the work along
  3. Check the result
  4. Decide what happens next

Watch a walkthrough

A walkthrough of BotSquad

An introduction to BotSquad and the idea of coordinating AI workers. The current interface uses tunnel-based browser access; standard web access and mobile interaction remain in development.

BotSquad

Give your AI workers a place to work together.

Loading this video connects to YouTube. Google's privacy policy applies.

Google privacy policyAbout website videos and privacy
Watch on YouTube

Why BotSquad

Tired of carrying context between chats?

One conversation has the plan. Another has the code. You’re copying between them and trying to remember what needs checking. BotSquad keeps named workers, tasks, messages and execution records in one place.

How it works

From an assigned task to a reviewed result.

Workers keep their names and saved context between tasks. They run when work is queued, with limits on how many can run at once. They are not continuously running models, and their memory is not unlimited.

  1. Assign a task

    Define roles and assign explicit tasks with clear boundaries. Workers keep their identities and can resume their conversations as work continues.

  2. Pass the work along

    Messages, task state and artifacts stay in the workspace. Workers can pass work along without making you carry every result between conversations.

  3. Check the result

    The current runtime executes work through Codex. Supported software workflows use scoped changes, recorded test results and independent review before integration.

  4. Decide what happens next

    Review outputs, pause work and decide on protected operations. Publication and infrastructure actions use specific approval paths, within the supported workflow.

An example task

Start with a question you need answered.

Compare two ways to build this small software project. Link the sources, explain the tradeoffs and flag what still needs checking. Don’t spend money, publish or contact anyone.
  1. Coordinate

    A manager assigns the question to a researcher.

  2. Investigate

    The researcher returns a written comparison, sources and unanswered questions.

  3. Review

    The manager checks the evidence and brings the result back to you.

Made-up task to show the workflow. This is not a customer result or a product screen.

Review and permissions

Read the work before you rely on it.

You can read tasks, messages, execution records and the files workers produce. Review matters: AI workers can make mistakes. Scoped permissions and approval steps help limit what they can do, but do not guarantee security or correct results.

Coordination state stays on your host. Configured external model services may receive task context for execution; self-hosted does not mean entirely offline inference.

Getting started

Trying BotSquad today.

Available now

Open source and self-hosted. Access the current interface in your browser through a tunnel.

MIT licensed. The public repository includes the source, setup instructions and validation records.

In development

Standard web access and a mobile app for interacting with BotSquad are in development. Neither is released, and no launch date is announced.

Standard web access does not imply a hosted SaaS service.

The current setup uses an operator-controlled Ubuntu host, Node.js and the Codex runtime. Follow the repository's setup instructions for supported versions and requirements.