Quickstart

This is the shortest path. Each step links to the full guide.

1. Deploy. From the product source checkout, scaffold a deployment directory, validate it and deploy.

   npm ci
   node cli/bin/cengrid.ts init deploy/layers/<org> --org <slug> --target <fly-or-aws>
   node cli/bin/cengrid.ts check --config deploy/layers/<org>/cengrid.config.jsonc

   Then run cengrid up --build-from <path-to-checkout> (add --yes on AWS) and cengrid check --live. See the deployment directory and targets.

2. Sign in as the first administrator with the email you put in ADMIN_GRANTS. See onboard your organisation.

3. Add a model provider in Admin → Onboarding → Model providers, and pick the base model in Governance → Models. See models.

4. Connect GitHub under Connections, so you can import plugins. See Connect GitHub.

5. Connect Slack (and optionally Google Chat). See Connect Slack.

6. Register Google under Connections → Available, and connect your own account from Keychain. See Connect Google Workspace.

7. Create a team and add a room under Admin → Rooms. See teams and rooms.

8. Add capabilities. Connect a plugin from GitHub and add its skills and instructions to a scope. See Connect capabilities from GitHub.

9. Invite your team and send a first message. See onboard people.

Prefer to delegate?

cengrid init writes deployment.md and a deploy-cengrid skill into your deployment directory. Hand them to a coding agent and it will run these steps for you, confirming account and billing choices with you before it changes anything.

Run it locally

For development, npm run dev runs the watch-and-reload server, and npm run dev-instance boots a production-shaped local instance reachable in Slack. Node is pinned in the product repository’s .node-version.