Connect your repo

Connect the repo your team works in and your machines get a lot more useful. They can look up code while you talk about a task, use commit history to answer questions, and make changes you review in a pull request.

An admin connects repos in Plugins → GitHub, under the Machines key in the dock. Start with the one behind your current project.

Let your commits move the board

Once GitHub is connected, your pushes and pull requests flow into WeMachines. Put a task code like T-42 in a commit or pull request and the work links itself to that task. Your board can then move on its own, say to In Review when a pull request opens and to Done when it merges.

You set those rules in Plugins → GitHub → your repo → Workflow. They read as plain sentences, so everyone on the team knows what starting work, pushing, opening a pull request or merging does to a task.

Ask questions about your code

Ask “where’s the retry logic?” or “how does our signup flow work?” and the machine searches your repos and answers with the files it read. It only reads, so a question can never change your code. You can switch code search off for any machine in its settings.

Got more than one repo?

Connect as many as you like. Each one has its own main branch, a switch for whether machines see its recent commits, and a short description of what’s inside. That description is how a machine knows where to look, so keep it clear, like “the marketing site and blog”.

Don’t feel like writing them? Press Study in Plugins → GitHub. It reads every repo, fills in the missing descriptions and draws a map of which repo uses which, each link backed by the file that proves it. It never overwrites what a person wrote, and it keeps going if you close the tab.

Hand a task to a machine

Every task has a play button: on its row, on its board card, and on the bar across the top of an open task. Press it, pick a model, press Run. The machine works on its own branch and opens a pull request. Or just say it: “@AI implement T-42”, in chat or out loud.

Sometimes a run needs your OK, for example to search the web. You’ll see Allow once, Allow for this run or Decline. Reading files, editing them and running commands are fine from the start. If nobody answers for ten minutes, the run stops and saves its work to the branch. Every run tells you why it ended, so nothing just goes quiet.

T-42 Fix the login redirect

Read 6 files

Edited 2 files

Asking to use web search It can't do this without you, and it's waiting. Allowing applies to this run only. {"query":"oauth redirect_uri trailing slash mismatch"} Allow once Allow for this run Decline

While your runs are going, a key in the dock counts them and turns red when one is waiting on you, so you can answer from anywhere in the app.

Let a machine answer the review

Every task with a pull request wears a small pill with its live state: open, merged, checks failing and so on. Review comments, from people or from bots like CodeRabbit, are listed on the task too. Tick the ones to fix and press Address. A machine works through them on the same branch, then replies in each thread on GitHub.

Bring the coding agents you already use

Claude Code, Cursor, Codex and friends can join your workspace too, over MCP. Run this one line in your repo’s root folder. You’ll also find it in Settings → Multiplayer:

It sets up four things, and it never overwrites anything you wrote. Files are only created if they’re missing, the two JSON files are merged, and running it again is safe.

Once connected, your agents can read and update tasks and docs, and their work shows up live next to your name and on the board. The guard script stops them writing to the board from outside your repo, and auto_mode in the config decides how freely they write: on, ask first, or only when you ask by name.

When you connect, you can also pick one of your machines for the agent to work as. Your terminal then gets that machine’s instructions, skills and plugins, just the way your team set it up.

Run on your computer or in the cloud

A run started in the browser happens in our sandbox and uses your workspace’s credits. Start one from the desktop app instead and it can use a coding agent already on your computer, like Claude Code, Codex, OpenCode, Copilot CLI or one of ten others, on your own plan and for no credits at all. It works in its own git worktree next to your checkout, so your files stay untouched and you can run several at once. Spend credits well covers when to pick which.

In the browser Press Run on a task

runs in

Our sandbox Its own branch, then a pull request

spends

Workspace credits Shared by the whole team

Desktop app Press Run on a task

runs on

That computer’s own agent Claude Code, Codex, OpenCode and more, in an isolated git worktree

costs

No credits at all That person’s own subscription