# Retell AI MCP

> Retell AI MCP connects your agent to a full telecommunications stack. It lets your AI client handle complex voice interactions, like initiating outbound calls or registering web-based sessions. You can build and manage entire AI personas, configure their underlying language models, and analyze call transcripts for sentiment analysis—all through simple chat commands.

## Overview
- **Category:** ai-frontier
- **Price:** Free
- **Tags:** voice-ai, conversational-agents, telephony, real-time-audio, voice-automation, customer-interaction

## Description

Your agent needs to talk on the phone? This MCP connects it directly to Retell AI’s voice platform. Instead of just chatting with text, your AI client can now orchestrate actual real-world calls. You can command your assistant to make an outbound call or set up a web browser session for testing. It’s like giving your chatbot a functioning phone line and a full operational team behind it.

Need to manage the agents themselves? You can dynamically create new AI personalities, adjusting their core instructions and models in one go. When the calls are done, you don't have to manually sift through records; you simply ask your agent to list all historical activity or pull detailed transcripts for deep analysis. Because this MCP lives on Vinkius, you connect once from any compatible client, giving your team access to managing voice agents and call telemetry without needing to log into a dozen separate dashboards.

## Tools

### create_agent
Creates and deploys a new AI voice agent personality.

### create_llm
Configures the specific language model that powers an agent's conversation brain.

### create_phone_call
Initiates a new outbound call by specifying the calling and receiving numbers.

### get_call_details
Retrieves comprehensive data for a specific phone call, including transcripts and sentiment scores.

### list_agents
Lists all AI agent profiles currently configured in the account.

### list_calls
Provides a log of both historical and active calls.

### list_phone_numbers
Shows all phone numbers associated with the account for calling.

### list_voices
Displays every available text-to-speech voice option.

### register_web_call
Sets up a new web browser session for call testing or integration.

### update_agent
Modifies the instructions or parameters of an existing AI voice agent.

## Prompt Examples

**Prompt:** 
```
Can you show me the transcripts for call ID `c_f3a123`?
```

**Response:** 
```
I accessed the tool `get_call_details` probing for the unique identifier `c_f3a123`. The call lasted 4 minutes and the sentiment analysis registered as 'positive'. Below is the full transcript of the conversation generated by Retell!
```

**Prompt:** 
```
List all available agents I can use.
```

**Response:** 
```
I sent a request evaluating all active profiles using `list_agents`. I discovered 3 agents in your account, including your 'Support Bot' (ID: agent_1), 'Sales Executive', and 'Reception Voice'. Let me know if you would like me to modify any of their instructions using `update_agent`.
```

**Prompt:** 
```
We are testing out new numbers. Please use 'from_number' `+18005551234` and dial `+14085551234` assigning my 'agent_555'.
```

**Response:** 
```
Done. I engaged `create_phone_call` injecting your JSON settings to ring out! Retell processing is initiated. Monitor your active calls using `list_calls` to see when it connects and check its metrics after completion.
```

## Capabilities

### Make Outbound Calls
Directly initiate phone calls using specified source and destination numbers.

### Manage AI Agent Personas
Create or update the specific instructions and models that define how an AI voice agent behaves.

### Record Web Calls for Testing
Register a new web-based call session, useful for testing browser integrations.

### Review Call History and Details
Pull logs of all past calls or retrieve the full transcript and sentiment analysis for one specific conversation.

### List Communication Assets
Check available phone numbers, text-to-speech voices, and configured agent profiles.

## Use Cases

### Quality Assurance Scripting
A QA engineer needs to test how the new support bot handles objections. They ask their agent to 'Register a web call and then initiate a mock objection handling flow.' The system uses `register_web_call` and simulates the full interaction, allowing immediate bug reporting.

### Post-Mortem Call Review
A sales manager needs to know why a deal stalled. They ask their agent, 'Show me the details for call ID 987.' The system uses `get_call_details`, providing the full transcript and confirming the sentiment score was negative.

### Scaling Support Coverage
The support team suddenly needs a specialized agent. Instead of spending an hour in a dashboard, they ask their agent to 'Create a new fraud detection agent.' The system uses `create_agent` and `create_llm`, deploying the persona instantly.

### Compliance Auditing
A compliance officer needs to audit all activity from last week. They ask their agent to 'List calls between Monday and Friday.' The system uses `list_calls` to pull the required log data for review.

## Benefits

- You don't need to write code to manage agents. Use the `create_agent` and `update_agent` tools to define new AI personalities and adjust their instructions via plain chat commands.
- Analyze call outcomes instantly. After a conversation, use `get_call_details` to pull transcripts and sentiment analysis without opening any separate dashboard tabs.
- Test web integrations easily. The `register_web_call` tool lets you set up virtual call sessions for debugging and quality assurance before going live.
- Control the entire communication stack. You can check all available assets—from calling numbers via `list_phone_numbers` to voice options using `list_voices`.
- Execute complex actions in one chat session. Combine tools like `create_agent`, `create_llm`, and `create_phone_call` into a single, natural language request.

## How It Works

The bottom line is that you chat naturally with your AI client, and it runs the complex telephony actions for you.

1. Install the Retell extension module into your MCP environment.
2. Get your API Key from your provider account and input it into the required authentication parameter.
3. Tell your assistant what you want to do, like: 'List all voices, create a support agent, and test a call to 555-1234.'

## Frequently Asked Questions

**How do I get an agent to make a call using Retell AI MCP?**
You use the `create_phone_call` tool by providing the required JSON payload, specifying both the calling number and the receiving number.

**Can I see what voices are available with Retell AI MCP?**
Yes, you call `list_voices`. This tool pulls a complete list of all text-to-speech options your account supports so you can pick the best one for your agent.

**What if I need to change an existing AI bot's behavior?**
You use `update_agent`. This tool lets you modify the instructions and parameters of a deployed agent without having to recreate it from scratch.

**How do I find out what calls happened yesterday with Retell AI MCP?**
Call `list_calls` first for a list of all call IDs. Then, use the specific ID you found with `get_call_details` to retrieve transcripts and sentiment.

**Does Retell AI MCP work for web browser testing?**
Yes, it does. You use `register_web_call` to establish a simulated web session, allowing you to test your agents in a live-like environment before deployment.