Endpoints, payloads, and the limits that apply.

The programmatic surface for teams that want conversation data in their own systems. Read transcripts and dispositions, push knowledge-base updates, subscribe to call-completed and escalation webhooks, and read agent configuration. Authentication is a scoped bearer token per environment.

The core resources

• Conversations — list, filter by disposition, and fetch a full transcript.
• Bookings — the appointment the agent created, with the source conversation attached.
• Knowledge — push or replace documents in the retrieval index.
• Agents — read configuration, escalation rules, and the active model chain.

Webhooks

Two events matter for most integrations: conversation completed, which carries the transcript, summary, and disposition tag, and escalation triggered, which fires the moment the agent hands off so your on-call person hears about it before the customer does. Payloads are signed and retried with exponential backoff for twenty-four hours.

Auth and limits

Scoped bearer tokens, separate per environment, rotatable without downtime. Rate limits are generous on reads and deliberately tight on knowledge writes, because re-indexing is expensive and a runaway sync is the most common way a customer degrades their own retrieval quality.