Skip to main content
Opengeni’s public API is a workspace-scoped HTTP API. Most integrations should use @opengeni/sdk, which wraps these routes with types, retries, and stream recovery. Every request authenticates first (see Authentication).

Billing

API and embedded-product usage spends credits, not plan credit. Connected AI subscriptions are available only in the Opengeni apps on Pro ($20/month), Max, or Enterprise; API requests that explicitly select one return 403 with subscription_api_unavailable. A self-hosted deployment can allow them for API requests with OPENGENI_API_CONNECTED_SUBSCRIPTIONS_ENABLED=true; see Self-host. See Going to production for billing and model selection.

Discovery

Main route families

Streaming

The event stream is Server-Sent Events. Every event has a sequence number, so a client can reconnect with the last sequence it saw and fill any gap from the replay route. Disable proxy buffering and set read timeouts of at least an hour.

Compatibility

The API only grows, and server updates don’t break existing integrations: servers accept every request a published SDK sends, and clients ignore unknown fields and event types. Rare exceptions (an emergency security fix or an explicit product-owner decision) are announced with migration notes. See Going to production. The full route list is in the repository’s HTTP API overview.