Quickstart

This path creates one working run from end to end. You need an owner or admin role, Docker running locally, and the devboxes command.

1. Get access

Accept your Devboxes invitation and sign in at app.devboxes.ai. The app stays locked until you choose Accept and continue on the beta terms.

If nobody has invited you, join the waitlist at devboxes.ai and confirm its email.

2. Connect a repository

Open Manage → Credentials → Repository and select Connect GitHub App. For GitLab, Gitea, Forgejo, or a self-hosted host, use Add manual PAT instead. See Repository connections for recovery paths.

3. Create a project

Open Projects → Create project, import the repository, choose a runner image preset, and finish. The first run stays Queued while its image builds. See Creating a project.

4. Start your runner

Run the hosted sequence without --api:

devboxes login
devboxes connect
devboxes credentials setup
devboxes doctor
devboxes listen

1. login approves this terminal in your browser and saves the account session.
2. connect registers this machine. It reuses the signed-in session when valid. Use --name <name> to choose the runner name.
3. credentials setup connects a provider subscription, accepts an API key, or saves a reference to a supported local auth file.
4. doctor checks registration, Docker, and credential readiness.
5. listen claims matching work and runs it in Docker.

listen --max-concurrent <count> raises the default concurrency of one. Self-hosted deployments set --api; hosted Devboxes is already the default.

5. Dispatch and steer

Open Agent sessions → Dispatch run. Choose the project, blueprint, model, branch, and task, then dispatch. Open the resulting agent session to watch the transcript, answer prompts, and review the diff.

To make a local credential available to dashboard-managed and other runners, run devboxes credentials sync. Managing credentials explains the expanded trust boundary.

Next: Devboxes CLI — dispatching and following runs from your terminal.