# Klatt

> Build, test, deploy, and operate voice workflows with Klatt.

## Docs

- [Inference providers](https://docs.klatt.ai/configurations/inference-providers.md): Choose the speech and language options available in Klatt, then verify them with a workflow test.
- [Turn interruption settings](https://docs.klatt.ai/configurations/interruption.md): Control whether a caller can interrupt an agent while it is speaking, and use that choice deliberately for each conversational step.
- [Tracing configuration](https://docs.klatt.ai/configurations/tracing.md): Connect approved tracing for an organization when you need runtime visibility into workflow behavior and provider timing.
- [Context and variables](https://docs.klatt.ai/core-concepts/context-and-variables.md): Use runtime context and template variables to pass known values into prompts, tools, and follow-up actions.
- [Getting started](https://docs.klatt.ai/getting-started.md): Create a first voice agent, test its draft behavior, and publish a version that is ready for real calls.
- [Klatt Documentation](https://docs.klatt.ai/index.md): Build, test, and operate voice workflows with Klatt.
- [MCP connections](https://docs.klatt.ai/integrations/mcp.md): Connect a customer-managed MCP server so its selected tools can be used by an agent with explicit authentication and filtering.
- [Custom SIP trunk setup](https://docs.klatt.ai/integrations/telephony/custom-sip-trunk.md): Connect a carrier- or PBX-managed SIP trunk to Klatt, then verify inbound or outbound calling before accepting real traffic.
- [Inbound telephony](https://docs.klatt.ai/integrations/telephony/inbound.md): Connect a Twilio number or custom SIP trunk to a Klatt workflow, then verify the call path before accepting real traffic.
- [Twilio setup](https://docs.klatt.ai/integrations/telephony/twilio.md): Configure Twilio credentials and numbers in Klatt, then verify the telephony path before accepting real traffic.
- [Add an agent to a website](https://docs.klatt.ai/voice-agent/add-to-website.md): Choose the widget mode that fits your page and configure the generated embed settings for a browser voice experience.
- [Agent node](https://docs.klatt.ai/voice-agent/agent.md): Use an Agent node for one focused conversational exchange with its own prompt, tools, documents, and optional extraction.
- [API Trigger node](https://docs.klatt.ai/voice-agent/api-trigger.md): Launch a published or draft workflow from a public HTTP trigger without turning this page into an API reference.
- [Editing a workflow](https://docs.klatt.ai/voice-agent/editing-a-workflow.md): Edit a draft safely, resolve node requirements, and verify changes before publishing them.
- [End Call node](https://docs.klatt.ai/voice-agent/end-call.md): Close a conversation cleanly, optionally capture final values, and end the call without leaving an open path.
- [Global node](https://docs.klatt.ai/voice-agent/global.md): Apply one shared persona, tone, or safety rule to selected prompted nodes without duplicating the text in every prompt.
- [Workflow fundamentals](https://docs.klatt.ai/voice-agent/introduction.md): Build a voice experience as a connected workflow whose nodes describe what happens before, during, and after a conversation.
- [Knowledge base](https://docs.klatt.ai/voice-agent/knowledge-base.md): Upload approved documents, wait for processing, and attach the resulting sources to an agent step for retrieval during a call.
- [Pre-call data fetch](https://docs.klatt.ai/voice-agent/pre-call-data-fetch.md): Fetch a small JSON context object before a call so the opening prompt can use current information.
- [Pre-recorded audio](https://docs.klatt.ai/voice-agent/pre-recorded-audio.md): Upload reusable audio, review its transcript, and select it for a greeting, goodbye, or tool message.
- [Start Call node](https://docs.klatt.ai/voice-agent/start-call.md): Configure the entry point that greets the caller, establishes initial context, and begins the first conversational turn.
- [Template variables](https://docs.klatt.ai/voice-agent/template-variables.md): Use the supported double-curly syntax to insert runtime values into prompts, request bodies, headers, and other configurable fields.
- [Call transfer tool](https://docs.klatt.ai/voice-agent/tools/call-transfer.md): Transfer a live call to a valid phone or SIP destination when the conversation requires a human or another service.
- [End Call tool](https://docs.klatt.ai/voice-agent/tools/end-call.md): Let the agent end a conversation when a defined condition is met, with an optional goodbye message and outcome reason.
- [HTTP tool configuration](https://docs.klatt.ai/voice-agent/tools/http-api.md): Call an external HTTP service from an agent step with explicit method, URL, authentication, parameters, headers, and timeout settings.
- [Tool lifecycle](https://docs.klatt.ai/voice-agent/tools/introduction.md): Give an agent a bounded action, configure its inputs and authentication, then verify its success and failure behavior in a test call.
- [Webhook node](https://docs.klatt.ai/voice-agent/webhook.md): Send a templated HTTP request after a workflow completes so another system can receive the call outcome.