Introducing the Vonage MCP Tooling Server

The Vonage MCP Tooling Server unlocks AI-driven automation for the Vonage platform, allowing developers to run SMS, voice, and WhatsApp operations directly from tools like Claude and Cursor.

Introduction: What If Claude Could Build and Manage Vonage Apps?

Following our recent announcement of the Vonage Documentation MCP Server, we’re continuing to expand how developers can use MCP to work with the Vonage platform. The documentation server brought official, up-to-date Vonage references directly into AI assistants and IDEs. Now, the new Vonage MCP Tooling Server takes the next step: instead of just retrieving information, your AI agents can perform real actions (sending messages, managing numbers, checking balances, and more) all from within the tools you already use.

A natural question that comes up as developers experiment with AI-driven workflows is whether their assistant can actually handle tasks, not just talk about them. Imagine prompting Claude to create a simple app that sends and receives WhatsApp messages, without any SDK setup or authentication boilerplate. That’s the idea behind the Tooling Server. And it works with any AI agent or chatbot that speaks MCP. Whether you are using Claude, Cursor, VS Code Copilot, Antigravity, or similar environments, you can drop the Tooling Server into your workflow. It’s stable, supported, and available today.

Why It Matters: Automate Communications With Dev Tools You Already Use

You’ll see the real value once it’s part of your workflow. AI agents can orchestrate tasks like “check the balance → send an SMS → save the result” without any custom scaffolding. Internal tooling gets easier because routine operations no longer require manual steps. The server enables your agents to provision numbers, send test messages, or verify connectivity automatically. And you can build AI copilots that actually perform actions instead of handing back code samples.

The Tooling Server and the Documentation MCP Server are designed to complement one another. The documentation server gives your agent authoritative answers about how Vonage APIs work. The Tooling Server enables the agent to act on that information. Together, they create a smoother, more integrated workflow for developers building with Vonage and experimenting with AI-native tooling.

What You Get Today: Tools Already Available

The initial release ships with a set of focused, single-purpose tools:

• balance — check your Vonage account balance

• sms — send SMS

• outbound-voice-message — send text-to-speech voice calls

• list-applications and list-numbers

• link-number-to-app

• WhatsApp and RCS support through the sendChannelMessage flow

Each tool is intentionally small, predictable, and structured in a way that works well for LLMs. Learn how to get started using these tools by adding Vonage APIs to Your AI Agent.

Getting Started: Simple, Local, and Secure

One important point about the design: the server doesn’t expose an HTTP endpoint. Instead, it runs as a local MCP server process: either inside your IDE, alongside a chatbot environment like Claude Desktop, or on any machine or container you choose to deploy and manage. In practice, this usually means running it via the CLI with your environment variables set, or installing it directly in your project as an NPM package through:

npm i @vonage/vonage-mcp-server-api-bindings

Here’s an example configuration block for using the Tooling Server with your IDE or Claude Desktop:

{
  "mcpServers": {
    "vonage-tooling": {
      "command": "npx",
      "args": ["-y", "@vonage/vonage-mcp-server-api-bindings"],
      "env": {
        "VONAGE_API_KEY": "...",
        "VONAGE_API_SECRET": "...",
        "VONAGE_APPLICATION_ID": "...",
        "VONAGE_PRIVATE_KEY64": "...",
        "VONAGE_VIRTUAL_NUMBER": "+1234567890"
      }
    }
  }
}

Note: Exact configuration depends on which IDE you use. 

Want More Tools? We’d Love Your PRs

Because the project is open-source, you can contribute your own tools or refine existing ones. Check out the CONTRIBUTING.md for more information. The structure is intentionally simple: one tool per file, Zod for schemas, and small, testable functions.

If there’s something you want to automate using Vonage APIs, this is a great place to start building it.

Conclusion

The Vonage MCP Tooling Server reflects a direction we’ve been moving toward for some time: developer workflows that are AI-native, agent-driven, and focused on concrete tasks rather than boilerplate. MCP is evolving quickly, and we want to help shape how it’s used in real communication systems, not just observe from the sidelines.

This release is an early step, but a meaningful one. The Tooling Server gives technical users and agent developers a foundation they can run locally or deploy on their own terms. At the same time, we recognize that not every builder wants to manage keys or operate infrastructure. To support a broader range of use cases, we’re also working on a hosted MCP server that brings the same capabilities into more accessible environments — from low-code platforms like n8n to internal automation tools.

Our goal is to provide a complete, flexible path for AI-powered communications. More tools, more coverage, and more ways to integrate with the workflows developers rely on. If you’re exploring what AI agents can accomplish beyond code generation, we’d love for you to build with us and help guide where this ecosystem goes next.