- sign in as a real AI Stats user from a local shell, remote shell, or agent runtime
- create and rotate API keys or management keys
- manage workspaces, presets, settings, and guardrails
- use first-party CLI OAuth while user-owned OAuth apps remain in private testing
- inspect models, providers, pricing, credits, activity, analytics, and generations
Install
Check the installed version
aistats --versionprints the installed version.aistats versionprints the version plus the recommended update command for the current package manager.- interactive runs also show an update hint when a newer published version is available.
Sign in
Start with:- Sign in with AI Stats: browser OAuth with authorization code + PKCE
- Sign in with Device Code: best for SSH, CI, headless shells, and agent environments
Confirm who you are
Create keys
--json when an agent or automation needs the raw created key once.
Manage workspaces and guardrails
OAuth apps
User-created OAuth apps are coming soon. This release keeps OAuth client creation in private testing while the first-party AI Stats CLI uses the shared OAuth/OIDC foundation for browser sign-in, device code, token refresh, revoke, userinfo, and JWKS/discovery flows. Self-hosted or preview deployments keep third-party OAuth disabled unlessAI_STATS_THIRD_PARTY_OAUTH_ENABLED is set to a truthy value. Accepted truthy values are 1, true, yes, and on, matched case-insensitively after trimming whitespace. Leave the variable unset while validating the first-party CLI path; enabling it exposes user-created OAuth app management and should only be paired with reviewed redirect URI, consent, client-secret, and workspace authorization configuration.
.env
Docker
Kubernetes
Security model
- CLI sessions prefer OS-backed secure storage where available.
aistats logoutrevokes the stored refresh token before clearing local state.- Management keys require explicit capabilities.
- Regular inference keys rely on guardrails, routing policies, and workspace settings rather than a separate scope model.