Call Settings

Call settings control how your agent manages the lifecycle of a phone call: when to end it, what to do when a caller goes silent, and how to transfer to a human when needed. Configuring these correctly ensures calls end gracefully, callers are never left in silence without a response, and escalations reach the right person.

Auto End Call

When Auto End Call is enabled, the agent can automatically hang up once the conversation is complete, for example, after the caller confirms their request has been handled, or says goodbye. Without this setting, the agent stays on the line indefinitely until the caller hangs up first.

End with Message

Toggle this on to have the agent speak a brief farewell before disconnecting. A default message (“Thank you for your call. Goodbye!”) is provided, but you can customise it to match your brand's tone.

End Call Prompt

The end-call prompt gives the agent its instructions for deciding when to hang up. Write it in plain English, for example:

End the call after the user says goodbye, confirms their request is complete, or asks to end the conversation. Always thank them before ending.

Be explicit about goodbye signals

List the specific phrases or situations that should trigger a hang-up. The more concrete the prompt, the less likely the agent is to end a call prematurely or to stay on the line awkwardly after the conversation is clearly over.

Silence Handling

Silence Handling controls what your agent does when a caller stops responding. By default, the agent re-prompts the caller after a period of inactivity, and can optionally end the call if the caller never responds. This replaces the previous fixed behaviour, which always repeated the same generic line three times then abruptly disconnected, with fully configurable, per-agent logic.

Enabled

The master toggle for Silence Handling. When turned off, the agent never auto-reprompts or auto-ends on silence; it simply waits for the caller indefinitely. Turn it on to activate the settings below.

Default behaviour

Silence Handling is on by default for all agents. Existing agents that have never been configured use the platform defaults (10-second timeout, 3 check-ins, 10 seconds between each), so there is no change in behaviour unless you explicitly adjust the settings.

Timing

First check-in after (seconds): how long the caller can be silent before the agent speaks its first check-in prompt. For most agents, 8–12 seconds works well. Very short values (under 5 seconds) can feel intrusive; very long values leave callers uncertain whether they are still connected.
Between check-ins (seconds): the gap between each subsequent check-in. Typically the same as, or slightly longer than, the first-check-in value.
Max check-ins: the maximum number of times the agent will check in before giving up. Set to 0 to disable check-ins entirely (the agent will only end the call if End call after check-ins is on and this is set to 0, so be careful).

Check-in Messages

Add one message per line. The agent speaks them in order: the first message on the first check-in, the second on the second, and so on. This lets you write escalating prompts, for example:

“Just checking you're still there; take your time.”
“I'm not able to hear you at the moment. Please feel free to call us back when you're ready.”

If you leave the list empty, the agent follows the Check-in Instructions field instead (see below), and the AI varies its wording naturally each time.

Use messages for branded scripts

Fixed messages are ideal for regulated industries or brand-sensitive agents where every word must be approved. Leave them empty and use instructions instead when you want a more conversational, varied tone.

Check-in Instructions

Shown only when no fixed messages are set. This is a short instruction to the AI describing how it should re-engage the caller, for example:

Politely check whether the caller is still there and offer to help. Use a warm, patient tone and vary your phrasing each time.

If left blank, the agent falls back to following any Silence Handling section you have written in your system prompt. This is the recommended approach for agents whose system prompts already include detailed silence instructions.

End Call After Check-ins

When enabled, the agent gracefully ends the call after exhausting all check-ins. If a Closing Message is provided, the agent speaks it before disconnecting, for example:

Please feel free to call us back when you're ready. Goodbye.

Leaving the closing message blank ends the call silently, which is fine for internal or automated flows but is generally not recommended for customer-facing agents.

Always set a closing message

A spoken closing line, even a short one, prevents the call from feeling like it was cut off. It reassures the caller that the line ended intentionally and tells them they can call back.

Call Transfer

Call Transfer allows the agent to hand the call to a human agent or another phone number. This is useful for escalations, complex enquiries, or any situation where a human touch is needed.

Warm transfer: the agent places the caller on hold (with hold music) and dials the destination. Once the human picks up, the agent privately briefs them with a short summary of what the call is about before connecting the two parties. The caller stays on hold for the whole briefing and is only brought into the conversation once it's finished, so the human is fully up to speed the moment they start talking. If the destination doesn't answer, the caller is taken off hold and told the transfer couldn't be completed.
Cold transfer: the agent immediately bridges the call to the destination number without holding the caller or briefing the recipient. Faster, but the human receives the call with no context unless the caller provides it themselves.
Transfer destination: the E.164-formatted phone number to transfer to (e.g. +441234567890).
Transfer prompt: instructions that tell the agent when to initiate a transfer, for example: “Transfer the call when the user requests to speak with a human, or when you cannot resolve their issue after two attempts.”

Warm Transfer Briefing

You control exactly what the agent says to the person it's transferring to. The briefing is resolved in priority order; the first option you set wins:

Fixed briefing: a set line spoken word-for-word, with no AI involved. Best when every word must be approved.
Briefing template: a line spoken word-for-word after placeholders are filled in. Supported placeholders are {agent_name}, {department} and {supervisor}. Any tag you mistype is left exactly as written, so a typo never breaks a live transfer.
Briefing instructions: guidance that shapes the AI-generated summary, such as its tone, length, or a detail it must always include (for example: “State the caller's account number and keep it under 12 words.”).
Nothing set: the agent falls back to its automatic summary of the conversation.

Set these once for the whole agent in the Call Transfer section (they appear when Warm Transfer is on). When you use Department Routing, any department can carry its own briefing that overrides the agent-wide one for that destination, useful when different teams need different hand-offs.

Templates vs instructions

Use a template when you need a predictable, scripted hand-off with a few dynamic details slotted in. Use instructions when you still want the AI to summarise the actual conversation, but on your terms: shorter, in a particular tone, or always mentioning a specific fact.

Transfer and Department Routing are mutually exclusive

Call Transfer and Department Routing (available in the Integrations section) cannot both be active at the same time. Enabling Department Routing disables the single-number transfer toggle. If you need to route to multiple destinations based on caller intent, use Department Routing instead.

Block Unknown Numbers

When enabled, the agent rejects inbound calls that arrive with an anonymous, withheld, or restricted caller ID. Outbound calls are unaffected. Use this to reduce nuisance calls on public-facing numbers.