Connecting Tuner MCP Server

Connect Tuner to your AI assistant using the Model Context Protocol (MCP). Manage your agent configurations directly from your AI tool using natural language.

The Tuner MCP Server is currently in Beta. Available actions and supported platforms may expand over time.

---

What You Can Do

The Tuner MCP Server gives your AI assistant full access to your workspace and Agent Settings configurations — both reading and writing.

Retrieve Information

Action	Description
List Workspace Agents	View all agents in your workspace along with their current status and configuration
Get Agent Details	Retrieve the full configuration of a specific agent, including its system prompt and settings
View Call Outcomes	List all call outcome definitions configured for an agent
View User Intents	List all user intent categories defined for an agent
View Behavior Checks	List all evaluation rules currently active on an agent
View Capabilities	List all allowed capabilities configured for an agent

Create & Update Configurations

Action	Description
Manage Call Outcomes	Create, update, or delete call outcome definitions
Manage User Intents	Define, modify, or remove user intent categories
Create Behavior Checks	Add new evaluation rules for agent behavior
Configure Capabilities	Set up or modify allowed capabilities for your agents

All actions performed through the MCP Server are applied to your Tuner workspace in real time. Changes made via your AI assistant are identical to changes made through the dashboard.

---

Prerequisites

Before connecting, make sure you have:

1. An active Tuner account — Sign up here if you haven’t already
2. A supported AI assistant with the required subscription:

Platform	Required Subscription
ChatGPT (chatgpt.com)	Plus or higher
Claude (claude.ai)	Pro or higher
Claude Code (Terminal)	Pro or higher
Cursor (cusor.com)	Any
Anti-Gravity	Any

---

Setup Guide

Choose your AI assistant below and follow the step-by-step instructions.

Claude

Connect Tuner to Claude

1. Open Connector Settings

   In Claude, go to Settings → Connectors.

2. Add Custom Connector

   Click Add Custom Connector and fill in the following:

   • Name: Tuner
   • URL: https://voice-api.staging.ginni.ai/mcp/

   Click Add.

3. Connect and Authenticate

   After adding the connector, click Connect. You will be redirected to Tuner to authenticate with your account. Sign in and authorize access.

4. Enable Tuner in Your Chat

   When starting a new chat, click the + button in the message input, then go to Connectors and make sure Tuner is toggled on.

You can verify the connection by asking Claude: “What agents do I have in Tuner?”

ChatGPT

Connect Tuner to ChatGPT

1. Enable Developer Mode

   In ChatGPT, go to Settings → Advanced settings and enable Developer mode.

2. Create a New App

   Once Developer mode is enabled, click Create app in the top right corner. Enter the Tuner MCP server URL:

   https://voice-api.staging.ginni.ai/mcp/

   Complete the setup and authorize access with your Tuner account.

3. Enable Tuner in Your Chat

   Start a new chat in Developer mode. Click the + button, then go to More and select Tuner to enable it for your session.

You can verify the connection by asking ChatGPT: “List my Tuner agents and their configurations.”

Tab

Tab

Tab

---

Authentication

The Tuner MCP Server uses OAuth for authentication. When you connect for the first time, you will be redirected to Tuner’s sign-in page to authorize access. Once authenticated, your session remains active and you won’t need to re-authenticate unless you revoke access.

For IDEs, The MCP server uses API-key authentication through an Authorization header. Get you API key from Tuner and use it asa bearer token when configuring the MCP in your IDE.

All actions performed through the MCP Server respect your existing workspace permissions. You can only access and modify agents within workspaces you have access to.

---

Example Prompts

Once connected, try these prompts with your AI assistant:

Create a new call outcome

“Add a new call outcome called ‘Appointment Booked’ with type Success. The definition should be: The customer successfully scheduled an appointment during the call.”

Add a behavior check

“Create a behavior check for my agent that verifies whether the agent confirmed the appointment date and time before ending the call.”

Define user intents

“Add the following user intents to my agent: Schedule Appointment, Reschedule Appointment, Cancel Appointment, and General Inquiry.”

Review current configuration

“Show me all the call outcomes and user intents currently configured for my agent.”

Get agent overview

“List all agents in my workspace and show me a summary of how each one is configured.”

Check behavior checks

“What behavior checks are currently active on my Customer Support agent? Show me the full list with their descriptions.”

---

Troubleshooting

Connection failed or timed out

Make sure you are using the correct MCP server URL: https://voice-api.staging.ginni.ai/mcp/. Verify that your Tuner account is active and that you have the required subscription on your AI platform.

Authentication error

Try disconnecting and reconnecting the Tuner connector. In Claude, go to Settings → Connectors and remove Tuner, then add it again. In ChatGPT, remove the app from Settings → Apps and re-create it.

Tuner not appearing in chat

Make sure you have enabled Tuner for your current chat session. In Claude, check Connectors from the + menu. In ChatGPT, check More from the + menu and ensure Tuner is selected.

Actions not working as expected

Ensure you are specifying the correct agent and workspace in your prompts. If the AI assistant is unsure which agent to modify, provide the agent name or ID explicitly in your message.

---

What’s Next?

Manage Tuner with MCP

Best practices and use cases for managing your agent with AI assistants.

Classifying Calls

Learn about call outcomes and user intents you can configure via MCP.

Creating Custom Evaluations

Understand behavior checks that you can create through your AI assistant.

Custom Integration

Send call data to Tuner via the REST API.