Skip to content

AI sessions

Bind a repo

Updated July 25, 2026

Binding tells Milestone which board a repository belongs to. Once bound, agent work in that repo lands on the right board, and your team's context for that board reaches your sessions.

Binding has two halves - one shared with your team, one on your computer. You need both, and it's easy to finish only the first.

Two halves, two places

Adding a repo to your workspace does not bind it on your machine. If the list still says no repos are bound after you've added one, you've done half. The second step is Bind on this machine.

Bind a repo

1

Open your repo settings

Go to Settings → Connections and find Repos. Bound repos, workspace repos, and local-only repos all appear in one list.

2

Add it to your workspace

Click Bind a repo, choose the repository and the boards it maps to, then Create binding. This is the half your team shares.

3

Bind on this machine

Find the repo in the list and click Bind on this machine, then pick the local folder. This is the half that makes it work on your computer.

Before you start

  • The repo needs a remote. Binding is keyed on the git remote URL, so a local-only repository can't be bound. The row will say so.
  • An admin creates the workspace binding. If you're not a workspace admin, ask one - you can still do Bind on this machine yourself afterwards.
  • The repo must come from a connected GitHub or GitLab integration. There's no way to type a remote URL by hand.

Pickup mode

Each binding has a Pickup mode for this binding:

  • Worktree - the default. Each task runs in its own isolated worktree on a feature branch, leaving your working copy untouched.
  • Direct - runs in your checkout on the current branch. Pushing is disabled.

If you don't bind a repo

Nothing breaks. Your agent works in a scratch folder and posts its results as comments instead of committing. What you lose is board-precise context and the ability to have work land in your actual codebase.

Related

Ready to try Milestone?

Give your team and its AI agents one shared memory.

Request early access