> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.jambonz.org/agent-studio/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.jambonz.org/_mcp/server.

# Agent Studio

## Docs

- [Agent Studio](https://docs.jambonz.org/agent-studio/overview.md): Build, test and run voice agents on your jambonz account, without writing a jambonz application.
- [Getting started](https://docs.jambonz.org/agent-studio/getting-started.md): Connect Agent Studio to your jambonz account, check your credentials, create an agent and call it from your browser.
- [Cascaded voice pipeline](https://docs.jambonz.org/agent-studio/agent-types/cascaded-pipeline.md): jambonz recognizes the caller's speech, a language model you choose writes the reply, and a voice you choose speaks it.
- [Realtime voice API](https://docs.jambonz.org/agent-studio/agent-types/realtime-voice-api.md): One vendor's model listens, thinks and speaks. jambonz carries the call, adds the transfer and hangup tools, and records every turn.
- [Connected agent](https://docs.jambonz.org/agent-studio/agent-types/connected-agent.md): An agent built and run on Dialogflow CES or ElevenLabs Agents, put on the phone by Studio.
- [Prompt](https://docs.jambonz.org/agent-studio/building/prompt.md): What the model is told before the call, the shared prompt library, per-call variables and guardrails.
- [Tools](https://docs.jambonz.org/agent-studio/building/tools.md): HTTP requests, MCP servers and the built-in tools an agent can use during a call.
- [Knowledge](https://docs.jambonz.org/agent-studio/building/knowledge.md): Give an agent documents and pages it can search during a call.
- [Conversation](https://docs.jambonz.org/agent-studio/building/conversation.md): How the agent answers, takes turns, handles interruptions and silence, and what is kept from each call.
- [Transfer & hangup](https://docs.jambonz.org/agent-studio/building/transfer-and-hangup.md): Hand the caller to a person, take them through a secure detour and back, and end calls cleanly.
- [Testing](https://docs.jambonz.org/agent-studio/testing.md): Call your agent from the browser or your phone and watch each turn as it happens.
- [Publishing](https://docs.jambonz.org/agent-studio/publishing.md): Edit a draft, publish it as a version, and put that version on the phone.
- [Calls and usage](https://docs.jambonz.org/agent-studio/calls.md): Every call your agents take, with its transcript, timings, tools, transfers and recording.
- [Experiments](https://docs.jambonz.org/agent-studio/experiments.md): Send a share of real calls to a second published version and compare how the calls go.
- [Result webhook](https://docs.jambonz.org/agent-studio/webhooks.md): Have each call's result posted to your own system when the call ends, signed so you can trust it.
- [Custom harness](https://docs.jambonz.org/agent-studio/custom-harness/overview.md): Your code decides what the agent says. jambonz handles the call, the listening and the speaking.
- [Custom harness quickstart](https://docs.jambonz.org/agent-studio/custom-harness/quickstart.md): From an empty folder to a test call with your own harness, in about 15 minutes.
- [Harness protocol reference](https://docs.jambonz.org/agent-studio/custom-harness/protocol.md): Every message between jambonz and a custom harness (protocol version 1).
- [@jambonz/sdk/harness reference](https://docs.jambonz.org/agent-studio/custom-harness/sdk.md): Build a custom harness in TypeScript or JavaScript without handling the protocol yourself.