How and why we built an MCP server for our Text-To-Speech

Learn how we built a TTS MCP server, enabling chatbot-driven integration of high‑quality voice synthesis and simplifying workflow automation for AI‑savvy users.

Overview

I will talk about building Hume AI’s TTS MCP server (in progress, should be released by May) that we provide to let AI-savvy users use our Text-To-Speech capabilities from a chatbot.

The themes are:

You can use technologies like MCP to serve (ai-savvy) users who have sophisticated workflows. Two years ago you would need to build bespoke UIs and APIs to model your users’ workflows, but now you can get 80% of the way there by just wrapping your core capabilities with something like MCP and let the chatbot do the rest.
AI Voice is finally here. It’s good enough to use in serious creative or professional projects.

Tech stack