Update Agent

Type for this agent

Title for Agent

Some description

Message users see when initially visiting a agent

The IANA (Internet Assigned Numbers Authority) formatted timezone your AI agent operates in.

Agent voice model to use

Agent language to be used for voice and transcriptions

Agent instructions for user interaction. This is for chatAgents only

Agent LLM model to use for its intelligence

Optional reasoning budget for models that support thinking. Use null to disable or low for short thinking.

Model to use for pauseAndThink function reasoning. Defaults to gpt-5.2.

Force post-conversation summary/reporting/spam detection to use a thinking-capable model when agentThinking is off and the selected model does not support thinking. Uses agentThinkingModel (default gpt-5.2).

(Requires Specific Paid Plan) The Speech-to-Speech Voice engine to use for voice, AI, and transcriptions

Speech to text engine for this agent

Enable Deepgram smart formatting for speech-to-text (punctuation + numerals stay enabled regardless). Default true.

DTMF your agent will send upon answering a call. Useful if you are forwarding calls to your agent from a platform that requires a DTMF verification before the call is connected.

The amount of time in milliseconds your Agent will wait before speaking the greeting (welcomeMessage).

Optional background audio to play on calls. Accepts a URL string or an object with url, loop (true or false), gain, and startAfterGreeting fields. Only HTTP(S) URLs are supported.

how many words a caller needs to speak before it is treated as an interruption

(in seconds) How long the agent needs to be speaking before it can be interrupted

(in milliseconds) the default amount of time between words before we consider caller speaking to be finished. Increasing this can help prevent unintentional interruptions for slow talkers.

(in milliseconds) the default amount of delay before processed speech is sent to the LLM for a response. Increasing this can help prevent unintentional interruptions for slow talkers. Note that this delay automatically increases and decreases depending on the number of interruptions over a period of time.

(in milliseconds) how much speech processing delay is added per interruption.

the maximum number of interruptions that will influence the speech processing delay.

(in seconds) how much silence (nothing said by the caller) before the agent asks are you still there?.

the maximum number of times your agent can ask are you there (callerResponseTimeout) before hanging up the call.

Whether the contact card for the caller should be injected into the agent context

Whether the call summary history for the caller should be injected into the agent context

Whether to automatically notify callers that recording is happening: true (string) means notify callers from area codes where it is legally required. false (string) means never notify callers. always (string) means notify every caller whether legally required or not that recording is in progress.

The phrase that will be spoken to the caller notifying them that recording is in progress: This call is being recorded is the default.

In seconds. How many seconds should the caller experience ringing before the agent answers. Valid value range is 1-60

When true, pre-answer functions return their raw results directly to agent context instead of sending those results back to the LLM for another processing step.

Optional JavaScript regex string. On inbound voice calls, if the inbound SIP CNAM matches this pattern, the call is immediately rejected before answer. Reject wins if both reject and transfer match.

Optional SIP reject code to use when cnamAutoRejectRegex matches. Defaults to 603 when omitted.

Optional JavaScript regex string. On inbound voice calls, if the inbound SIP CNAM matches this pattern and reject did not match, the call is immediately transferred silently.

Destination used when cnamAutoTransferRegex matches. Accepts E.164 phone numbers or PBX/extension targets like ext:1234 or pbx:1234.

Require callee to press 1 before a transfer is completed

Continue to monitor and transcribe transferred calls.

Your agent will stay on the line with the caller while transfer attempts are made. Those answering the transfer attempts will need to say Yes to accept the call. This will override the confirmTransfers option.

What is required to be spoken by the transferee in order for the call to be bridged. Values can be null (defaults to Yes), any, or a regex.

The length of time in seconds we will wait for confirmation after a transferred call with confirmTransfers or attendedTransfers is answered.