Bring Vapi Voice Agents into Your Workflows With The New Vapi MCP Server

Today, we’re launching the Vapi Model Context Protocol (MCP) Server - a new way to control your voice agents using function calling from any ai agent tool.

With this integration, you can now ask your agent to spin up real-time voice flows, trigger outbound calls, or route to Vapi assistants-without leaving the chat interface. It’s a lightweight but powerful bridge between natural language input and real-world phone infrastructure.

Why It Matters

Bringing LLMs and automation closer together unlocks some serious workflow efficiency. Here’s what you can now do in seconds:

Trigger outbound calls directly from Claude desktop
“Call the customer back with the Riley assistant and confirm delivery for order #3478.”
Schedule follow-ups without context switching
“Book a call with the assistant tomorrow at 2pm using my phone number.”

Instead of jumping between dashboards, SDKs, or cron jobs, you can now use plain language to control live phone flows.

What’s New

The MCP server supports both local and remote setups:

Claude Desktop Integration
Configure Claude to use the Vapi MCP server and trigger voice assistant in real time.
Remote Access via SSE
Connect to https://mcp.vapi.ai/sse to access Vapi functionality from any MCP client - no local server required.
Function Calling Support
Claude Desktop can now pass structured commands directly to Vapi’s API, letting you invoke assistants, transfer calls, and more.

Real-World Examples

Teams are already using the MCP server to power high-leverage workflows:

Healthcare

A nurse uses Claude to say: “Can you use vapi assistant to call the patient to confirm their 10:30am appointment tomorrow?”
The MCP server picks it up, and Vapi routes the call to a scheduling assistant with the right context.

Sales

“Hey Vapi, use the SDR assistant to call Mary on +1234567890. Confirm she received the pricing sheet, ask if she has questions, and offer to schedule a follow-up.”
Claude handles the language, Vapi makes the call.

Support

“Cancel the pizza order, then ask if they want to speak to a human”
The assistant can detect the cancellation and decide to escalate - no custom triage logic required.

Getting Started

Open Claude Desktop → Settings → Developer → Edit Config
Add your Vapi API key and enable the @vapi-ai/mcp-server
Restart Claude Desktop
Create a phone number and assistant in your Vapi Dashboard
Ask Claude to trigger a voice workflow - the server takes care of the rest

You can also run the MCP server locally, or use our hosted SSE endpoint for remote calls.

What’s Next

We see MCP as the connective tissue for a lot more to come:

Multi-turn assistant control via Claude
Combining tool-calling with Claude’s long-term memory
Seamless agent orchestration between chat and voice

For now, this is the first step-and it’s already powerful.

Try it today. Configure your Claude Desktop or Cursor Agent, connect a Vapi assistant, and ask it to make a call. Your workflows just got a voice.