Agents, voice calls, and chats

Configure an agent

List your workspaces, then inspect available voices and LLM configurations. Create or update an agent with the supported response-engine configuration. Conversation-flow agents use the corresponding flow resource. Read the generated schemas for each engine variant.

Read an existing configuration before making a partial update. Publishing a version changes the configuration used by subsequent calls; verify the workspace and draft version first.

Start a voice call

Use POST /api/v1/calls/initiate with the agent, source number, and destination specified by its request schema. This places a real call and may incur charges. Use a dedicated test agent and a destination you control when validating an integration.

Read the resulting call with the call-detail endpoint or use the call-history endpoints. A successful creation response is not proof of completed audio or the desired conversation outcome.

For browser voice sessions, use POST /api/v1/calls/web and handle the returned session credentials as sensitive, short-lived access information.

Chat lifecycle

Create a chat with a chat agent, send messages through the chat-completions endpoint, inspect the session, and end it when the workflow is complete. Outbound SMS can contact a real recipient; confirm your test target before sending.

Phone assignments

Manage number assignments in the dashboard. The current public contract does not expose the dashboard-only number-management routes. Inbound schedules and default agents remain independent of outbound assignments.