Katexs
Tools overview

Tools

Tools overview

Give agents the ability to act: check calendars, book appointments, transfer calls and call your own APIs, within strict latency budgets.

What tools are

A tool is something an agent can do in the middle of a conversation. The model decides when to call a tool based on its description and your prompt, Katexs (or your endpoint) runs it, and the result is fed back so the agent can speak or write the answer.

Voice pipeline and latency targets Caller speech is transcribed, reasoned over by the model and synthesized back to audio. A tool call branches off the model and must answer within 800 milliseconds or acknowledge asynchronously. Callerspeaks TranscriberSpeech to texte.g. Deepgram ModelPrompt, memory, toolse.g. GPT-4o, Claude VoiceText to speeche.g. Cartesia Callerhears Tool callAnswer in <800 ms or ack async Your endpointPOST tool-calls Latency targets per turn <200 ms <300 ms to first token <150 ms to first audio
The voice pipeline for one conversational turn, with the latency target for each stage. Targets, not guarantees: provider and network conditions vary.

Types of tools

ToolWhat it doesWhere it runsStatus
Calendar toolsCheck availability, book, reschedule and cancel appointments on Google CalendarKatexsBeta
Transfer callForwards the caller to a phone number you chooseVoice engineLive
End callEnds the call when the agent says an end-call phraseVoice engineLive
Custom functionsCalls your HTTPS endpoint with arguments the model fills from a JSON schemaYour serverLive
Knowledge searchLooks up answers in the agent's knowledge baseKatexsLive

How a tool call works

  1. 1

    The model emits a tool call with a name and JSON arguments.

  2. 2

    Katexs checks the tool is enabled for this agent, applies rate limits and validates the arguments against the tool's schema.

  3. 3

    The tool runs. Voice tools have an 800 ms budget.

  4. 4

    The result comes back as one short, speakable string. Failures come back as calm messages the agent can say, never as stack traces.

  5. 5

    The call is logged in tool_calls with redacted input and output, its status and its latency, and appears in Traces.

Note

Tools are permissioned per agent. A tool that exists in your workspace does nothing until you enable it on an agent.

Rate limits

ScopeLimit
Per workspace (all tools and functions)30 calls per minute
Per contact (built-in tools)10 calls per minute

When a limit is hit the agent hears "I'm doing a little too much at once. Give me a moment and try again." and the call is logged as rate_limited.

In this section

Last updated November 2026 · Found a problem? Email support@katexs.com