Skip to main content
Checkfu is in private alpha. There is currently no self-service public CLI installation channel. The CLI artifact is not published for general download, and these docs do not provide an installation command. Every checkfu … example in these docs still describes the real command surface. Commands assume an explicitly authorized alpha arrangement. Anything that looks like a Checkfu CLI package on a public registry is not ours.
Plain HTTP against the Checkfu API needs no private package. The Checkfu CLI and TypeScript SDK are optional private-alpha artifacts; the HTTP requests work without them. A Checkfu source checkout is for contributors and authorized evaluators; customers do not need one for this page. Part of this flow is behind a rollout fence; a deployment that has not enabled it refuses the request rather than running it. Confirm capability status before making availability part of your application’s contract.

Paths that work today

  • Invited evaluators: use the artifact you received. It is the same command surface these docs describe.
  • Contributors and local evaluators: work from a Checkfu source checkout, where pnpm exec checkfu runs the CLI directly.
  • Everyone else: call the HTTP contract from your backend. Start with the HTTP API reference; nothing there needs the CLI.
checkfu connect, and therefore MCP setup, becomes available once you have CLI access through one of the first two paths. For a Runner-based Session, checkfu doctor --agent <agent-id> --environment <environment-id> checks API access and its declared wire version, the Workspace’s visible Runner presences, those two resources, and the Workspace’s public Session capability status. An empty or cold Workspace Runner list is informational: it does not expose hosted platform capacity, which is checked at Run start. A nonzero result names a missing public prerequisite or incomplete qualifying release evidence. A Session support-readiness failure is distinct from admission for an exact runtime tuple; a successful Session does not promote the Workspace’s support status. A successful result is still a public preflight: Session creation decides admission for the exact HarnessProfile, model route, permissions, and placement.