Skip to main content
Tuner is an observability, evaluation and simulation platform for voice AI agents. You send it your calls. It scores each call against evals you define, flags the calls that went wrong, and places test calls against your agent before and after you change it.

​
How Tuner fits into your workflow

1

Simulate

Test your agent with Call Simulation before launch and after every change. The evals that score production calls also score every simulated call.
2

Observe

Every call arrives with its transcript, per-turn timing, tool calls and eval results. Open a call in Call Logs to see what happened.
3

Monitor

Track performance on the Overview dashboard, and get alerts when a condition you define is met.
4

Improve

Fix the agent, check the fix against production data, and simulate again to catch regressions. Diagnose with MCP looks for root causes across every affected call.
The Overview dashboard with headline metrics, intent and outcome breakdowns, and recent red flags

​
Three ways to connect

Retell, Vapi, ElevenLabs and Dograh

Guided setup in the dashboard. Retell, Vapi and ElevenLabs connect with your platform key; Dograh uses the Tuner node in your workflow.

SDKs

LiveKit, Pipecat, or your own Python stack. The SDK records per-turn timing and latency for each pipeline stage.

Direct API

Any language. You build the call payload and send it after the call ends.
The SDKs record what happened turn by turn: end-of-utterance delay, time to first token and time to first audio byte. A platform that hands over a finished transcript can show that a call was slow; SDK data shows which pipeline stage was slow. Every SDK and API integration needs a Tuner API key, your workspace ID and the agent’s Agent Remote ID. See API keys and IDs.

​
Core concepts

Each concept links to the page that covers it. Every other term is in the Glossary.

​
Next steps

Quickstart

Create an agent and get your first call evaluated.

API keys and IDs

The three values every SDK, API and MCP integration needs.