Create an AI phone agent
Create an AI phone agent powered by the latest voice model. Use when orchestrating voice calls programmatically — placing outbound calls or listing call history.
Authorizations
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Body
Agent display name
Agent type: "inbound" (receives calls, default) or "outbound" (makes calls)
inbound, outbound Voice ID from the /phone/voices endpoint. Omit for default voice.
ISO-639-1 language codes the agent should speak. A subset of {en, es, fr, de, hi, ru, pt, ja, it, nl} can be combined freely. Any other code (e.g. "ar", "tr", "zh") must be the only entry; Arabic accepts dialect tags such as "ar-SA".
1What the agent says when it picks up (default: "Hello?")
The call prompt as an array of instruction blocks. Joined in order into the system prompt at call time. Sending this on update replaces the whole array.
HTTP actions the agent can call as tools during calls (action IDs from GET /actions). The full set is replaced on every write; send [] to detach all, omit or send null to leave it unchanged. IDs must belong to your organization.
Transfer destinations the agent may transfer calls to (IDs from GET /phone/transfer-destinations). The full set is replaced on every write; send [] to detach all. Destinations themselves are managed in the dashboard.
^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$Destination the call is transferred to if the voice pipeline fails mid-call. Any destination of your organization; null falls back to the first allowed destination, then to a graceful hangup.
^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$Speaking speed on a 1-100 scale (null = the voice default). Lower is slower; natural-sounding voices in some languages (e.g. Dutch) run best around 80-85.
1 <= x <= 100Custom vocabulary to boost in speech recognition: brand and product names, jargon, spellings the transcriber typically mishears. The agent name and transfer destination names are merged in automatically.
Record calls handled by this agent (default: false).
Response
Default Response
inbound, outbound oppie-vox, oppie-vox-turbo, oppie-vox-livekit office, bar, city-street deepgram/flux-general, deepgram/nova-3, deepgram/nova-3-medical, cartesia/ink-whisper, elevenlabs/scribe_v2_realtime, xai/stt-1, assemblyai/universal-3-5-pro, speechmatics/enhanced, soniox/stt-rt-v5 -9007199254740991 <= x <= 9007199254740991^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$Transfer destinations the agent may transfer to (transfer_call targets).
^([0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[1-8][0-9a-fA-F]{3}-[89abAB][0-9a-fA-F]{3}-[0-9a-fA-F]{12}|00000000-0000-0000-0000-000000000000|ffffffff-ffff-ffff-ffff-ffffffffffff)$