Quickstart

From a repo to a live URL. The first step happens in the browser because granting GitHub access is a browser permission — there is no token that skips it. Everything after that you can do from your agent, the dashboard, or CI.

1. Connect a repo

• Sign in at dash.primdb.com.
• New app → install the PrimDB GitHub App on the account or org that owns your repo.
• Pick the repo. PrimDB detects the build — see what it can build.

That is the whole browser part, and it is only needed for GitHub. If your code lives on GitLab, Bitbucket, Gitea or any https git host, skip it: point PrimDB straight at the remote with set_git_source over MCP. Those hosts have no push webhook here, so their deploys are manual — deploy_app, or the CLI from your CI.

2. Hand it to your agent

This is the shortest path and the one PrimDB is built for: nothing to install, and your agent gets all 110 tools rather than a subset. Create a token in Settings → API tokens, then point your MCP client at the endpoint.

• Endpoint: https://mcp.primdb.com
• Auth: Authorization: Bearer primdb_...
• Setup per client: MCP

Then say what you want: *deploy my app, then tell me if the build failed and why*. Your agent can also add a database, set env vars, wire a domain and read the deploy logs back to you — see /agents for a prompt you can paste.

Or click Deploy

The dashboard does everything the agent tools do. If you just want the app live, press Deploy on the service and watch the log stream.

Deploying from CI

For a pipeline you want a fixed command and an exit code, not a model. That is what the CLI is for: PRIMDB_TOKEN=... npx @primdb/cli deploy --wait fails the build when the deploy fails. See CLI.

Honest note. Every command on this page runs today. Where PrimDB cannot do something yet — non-GitHub repos, primdb db, domains from the CLI — the docs say so rather than showing you a command that does not exist.