Tools overview

Give your agent built-in actions and custom API or client tools it can call mid-conversation.

View as Markdown

The Tools tab (left sidebar, in the agent editor) is where you add capabilities like hanging up, transferring calls, knowledge base search, or API calls. Each tool has a Name and Description, which the model uses to decide when to use it.

Tools tab in the agent editor showing system tools panel and custom tools list
The Tools tab, showing System tools on the right and custom tools on the left

System Tools

Built-in actions, each is just a toggle, switching one on for the first time opens its configuration form.

Tells the agent when to hang up.

FieldDescription
NameInternal identifier for the function
DescriptionWhat the AI reads to decide when to trigger it

Be specific, vague descriptions cause premature hang-ups. “End when the customer confirms their issue is resolved” works better than “end the call.”

Hands the conversation off to a human.

FieldDescription
Name / DescriptionSame as above
Transfer NumberThe destination phone number
TypeCold Transfer (handed off immediately, no debrief) or Warm Transfer (the AI briefs the next agent first)

Warm Transfer adds three more sections:

  • During Transfer Call — pick On-hold Music (Ringtone, Relaxing Sound, Uplifting Beats, or None).
  • During Agent Connection — a Whisper Message: spoken only to the human receiving the transfer, before they’re connected to the caller. Written as a prompt or a static sentence.
  • After Transfer Connects — a Three-way Message: spoken to both parties once the transfer connects.

“Transfer only if a human answers” is visible here but marked Coming Soon, it can’t be enabled yet.

Lets the agent search a knowledge base for relevant context mid-call.

FieldDescription
Name / DescriptionSame as above
Knowledge BasePick from your existing knowledge bases
Filler PhrasesComma-separated phrases the agent can say while the search runs

The knowledge base itself has to already exist, this tool just attaches one. Create and populate one from the Knowledge Base section first.

Custom Tools

For anything not covered by a system tool, connect your own API or app logic. Click Add tool and choose a type:

Tool API Calls
The Tools tab, showing System tools on the right and custom tools on the left

Calls your API during the conversation and feeds the response back to the agent.

FieldDescription
Name / DescriptionSame pattern, the model reads the description to decide when to call it
LLM ParametersValues the model extracts from the conversation, referenced elsewhere as {{parameter_name}}. Each has a name, type (Text/Number/Boolean/Enum), description, and required flag
URL / MethodEndpoint and HTTP method (GET/POST/PUT/PATCH/DELETE)
TimeoutHow long to wait for a response, default 5000ms (range 1000-30000ms)
HeadersKey/value pairs
Query ParametersKey/value pairs
Request BodyJSON, for POST/PUT/PATCH only, supports {{param}} templating
Response Variable ExtractionPull fields out of the response into variables, by name and JSON path (e.g. $.data.user.name)
Tool API Calls
The Tools tab, showing System tools on the right and custom tools on the left

Runs inside your own app instead of hitting an API, useful when the “tool” is really a function in your frontend (e.g. update a cart, navigate a screen).

Client tools only work on websocket sessions (webcall, widget). Phone calls don’t support them.

FieldDescription
Name / DescriptionThe model reads the description to decide when to call it
ParametersSame shape as an API tool’s LLM Parameters, typed arguments your app receives
Wait for responseIf on, the conversation pauses until your app returns a result. Reveals a timeout slider (1-60s, default 1s)
Filler PhrasesOne per line, spoken while the tool runs so the pause isn’t silent