RetellAI MCP Server

The RetellAI Model Context Protocol (MCP) server bridges AI assistants with the full range of RetellAI’s voice‑centric communication services. By exposing a set of well‑defined tools, the server lets assistants create and control phone calls, manage voice agents, provision phone numbers, and select from a catalog of synthetic voices—all through the MCP interface. This enables developers to embed real‑time voice interactions directly into conversational agents without handling low‑level telephony APIs.

Problem Solved

Traditional voice integration requires developers to write custom SDKs, manage authentication, and orchestrate call flows manually. For AI assistants that aim to initiate or participate in phone conversations, this complexity hampers rapid experimentation and deployment. The RetellAI MCP server removes that barrier by providing a single, consistent protocol endpoint where assistants can issue declarative commands such as “create a phone call” or “list my agents.” The server translates these high‑level actions into RetellAI’s underlying telephony and voice services, returning structured responses that the assistant can immediately consume.

What It Does

When an MCP client connects, it discovers four categories of tools:

• Call Tools – Create, list, retrieve, or delete phone and web calls.
• Agent Tools – Manage voice agents that encapsulate LLM configurations, including creation, updates, and versioning.
• Phone Number Tools – Provision, configure, and query phone numbers that can be used as caller IDs or reception points.
• Voice Tools – Enumerate and inspect synthetic voices available for use in calls.

Each tool is a declarative operation; the assistant supplies the required parameters, and the server returns a JSON payload with status, identifiers, or detailed metadata. This design keeps interactions stateless and easily testable.

Key Features & Capabilities

• Unified Voice Management – One API surface for all voice‑related actions, from agent creation to phone number provisioning.
• Declarative Call Flow – Initiate outbound calls or web calls with a single tool invocation, specifying caller identity and target number.
• Agent Versioning – Retrieve historical versions of an agent, enabling rollback or audit trails.
• Rich Metadata – Every tool returns contextual information (e.g., call duration, agent configuration), allowing assistants to adapt responses dynamically.
• Extensible Tool Set – The server’s architecture supports adding new tools (e.g., voicemail handling) without changing the MCP contract.

Use Cases & Real‑World Scenarios

• Automated Customer Support – An assistant can dial a customer, route the call through a pre‑configured agent that handles FAQs, and log the interaction for compliance.
• Remote Order Fulfillment – As shown in the example use case, an assistant can create a pizza‑ordering agent, place a call to a local restaurant, and provide payment details—all orchestrated via MCP commands.
• Voice‑Enabled Surveys – Deploy agents that call respondents, gather answers through speech recognition, and store results for analytics.
• Multi‑Channel Communication – Switch between phone calls and web calls (e.g., video chat) using the same tool set, simplifying hybrid workflows.

Integration with AI Workflows

Developers embed the MCP server into their existing Claude Desktop or other MCP‑compatible assistants by adding a single configuration entry. Once registered, the assistant can invoke any of the available tools as part of its prompt generation pipeline. Because MCP is stateless, assistants can chain calls: list agents → create a new agent → initiate a call. The assistant’s natural language understanding can generate tool arguments on the fly, making voice interactions feel seamless to end users.

Unique Advantages

• Zero Telephony Overhead – No need to manage SIP, TURN servers, or telephony infrastructure; RetellAI handles all provisioning.
• LLM‑Powered Agents – Each agent can be tuned with different LLM configurations, allowing a single server to support diverse conversational styles.
• Scalable Voice Catalog – Access a growing library of voices, including multilingual options, without additional licensing headaches.
• Developer‑Friendly Toolset – Clear, self‑describing tools reduce the learning curve for developers familiar with MCP but new to voice services.

In summary, the RetellAI MCP server turns complex telephony and voice orchestration into a set of simple, declarative tools that AI assistants can consume directly. This empowers developers to add real‑time voice capabilities to conversational agents quickly, reliably, and with minimal infrastructure overhead.