Verbatik AI
MCP Server

Give Your AI Agent a Voice

Connect Claude, Cursor, Lovable, Windsurf, Codex, OpenAI, Kiro, or any MCP-compatible assistant to Verbatik — and let it speak. 2,700+ voices. Instant voice cloning. Zero code.

Connect Now — It's Free to Set UpView Documentation

Works with Claude Desktop, Cursor, Kiro, and any MCP-compatible client

Your AI Client

Claude Desktop
Cursor
Kiro
Windsurf
Any MCP Client
MCP Protocol

Verbatik Tools

2,700+ TTS Voices
Voice Cloning
Emotion Control
50+ Languages
Cost Estimation

What Is the Verbatik MCP Server?

The Model Context Protocol (MCP) is an open standard that lets AI assistants use external tools — the same way a browser uses APIs. Verbatik's MCP server turns your AI assistant into a full text-to-speech and voice cloning studio.

Instead of switching between apps, copying text, pasting into a TTS tool, downloading audio, and uploading it somewhere else — you just ask your AI assistant. It handles everything.

“Generate an audio version of this blog post in a French female voice.”

Done. One sentence, one audio file, one URL.

2,700+ Voices

Neural TTS in 50+ languages

Voice Cloning

Clone from 10s of audio

Zero Code

Plain language commands

Open Standard

Works with any MCP client

How It Works

Three steps to give your AI assistant a voice. No coding required.

1

Connect (30 seconds)
Open your AI client's MCP settings. Add the Verbatik endpoint. Authenticate with your API key or one-click OAuth. That's it.

2

Ask in Plain Language
No API calls. No code. Just tell your assistant what you need — "Read this paragraph aloud using the Jenny voice" or "Clone my voice from this audio file."

3

Get Audio Instantly
Your assistant calls the right Verbatik tools, generates the audio, and returns a shareable URL. Stream it, download it, embed it — your call.

Built for Agentic Workflows

MCP isn't just a shortcut — it's a paradigm shift. When your AI assistant has direct access to Verbatik's tools, it can chain actions together autonomously.

Content-to-Audio Pipeline
Your agent reads a blog post, picks the right voice for the audience, generates speech, and returns the audio URL — all from a single prompt.
Multilingual Voiceover
"Translate this script to French, Spanish, and Japanese, then generate audio for each language with a native voice." The agent handles everything.
Voice Cloning + Production
Upload a voice sample. The agent clones it, then uses the clone to narrate an entire document with emotion controls — happy for the intro, neutral for the body.
Cost-Aware Batch Processing
Before processing 50 articles, the agent checks your balance, estimates the total cost, and either proceeds or warns you to top up. No surprise charges.
Voice Library Management
"Show me all my cloned voices, delete the ones I haven't used, and generate a preview with the remaining ones." The agent audits your voice library.

9 Tools Your AI Assistant Can Use

Once connected, your AI assistant automatically discovers all available tools. No configuration needed — just ask.

Text-to-Speech
list_voices

Browse 2,700+ pre-trained voices. Filter by language, gender, or name.

text_to_speech

Convert up to 50,000 characters to speech. Automatic chunking. SSML support. Returns a stored audio URL.

cloned_voice_tts

Generate speech with your cloned voice. Control emotion, speed, pitch, volume. Supports interjections like (laughs), (sighs), and custom pauses.

Voice Cloning
clone_voice

Clone any voice from a 10-second audio sample. MP3 or WAV. Noise reduction included.

list_my_voices

See all your cloned voices with status, IDs, and preview URLs.

get_my_voice

Get full details on a specific clone.

delete_my_voice

Remove a cloned voice permanently.

Account & Billing
get_balance

Check your balance, monthly spend, and billing summary before generating.

estimate_cost

Know exactly what a job will cost before you run it. No surprises.

Works With Every Major AI Client

Verbatik's MCP server uses the Streamable HTTP transport — the most widely supported MCP protocol. Connect from any of these clients in under a minute.

Claude Desktop
Recommended
One-click OAuth or config file. The easiest setup.
Claude Code
Add to your settings.json. Native HTTP support.
Kiro
Drop the config into .kiro/settings/mcp.json.
Cursor
Add to .cursor/mcp.json. Works immediately.
Any MCP Client
If it supports Streamable HTTP, it works. For stdio-only clients, use the mcp-remote bridge.

Who Uses This

Podcasters & Content Creators
Turn show notes into audio intros. Generate episode teasers in your own cloned voice. Produce multilingual versions of your content without re-recording.
Developers & DevOps Teams
Add TTS to CI/CD pipelines. Generate audio alerts, status updates, or user-facing voice content programmatically through your AI coding assistant.
Educators & Course Builders
Convert lesson plans to audio. Generate voiceovers for slides. Create accessible audio versions of written materials in 50+ languages.
Product Teams
Prototype voice interfaces. Generate placeholder audio for mockups. Test different voices and emotions before committing to production.
Accessibility Teams
Generate audio versions of documentation, help articles, and UI text. Support users who prefer or require audio content.

What You Can Do With Verbatik Voices

Pre-trained Voices
  • 2,700+ neural voices across 50+ languages
  • Male, female, and neutral options
  • Multiple speaking styles per voice
  • SSML markup for pronunciation, pauses, and emphasis
Cloned Voices

Clone any voice from a 10-second audio sample. Once cloned, you get full control:

7 emotions: happy, sad, angry, fearful, disgusted, surprised, neutral
Speed: 0.5x to 2x
Pitch: -12 to +12 semitones
Volume: 0 to 10
Voice modification: pitch, intensity, and timbre adjustments (-100 to +100)
Natural interjections: (laughs), (sighs), (coughs), (gasps), (yawns), and more
Custom pauses: 0.01 to 99.99 seconds
Language boost: 30+ language options for multilingual clones
Output formats: MP3, PCM, FLAC with configurable sample rate and bitrate

Simple, Transparent Pricing

No subscriptions. No monthly fees. No per-seat charges. Just pay for what you generate.

WhatCost
Pre-trained voice TTS$0.002 per 1,000 characters
Cloned voice TTS$0.10 per 1,000 characters
Voice cloning$3.00 per clone
Browsing voices, checking balance, estimating costsFree

MCP uses the same prepaid balance as the REST API. Top up once, use it everywhere — dashboard, API, or MCP.

Your AI assistant can check your balance and estimate costs before generating.

Secure by Default

Two authentication methods. Pick what works for you. Both use the same prepaid balance. Both work with all 9 tools. Create separate API keys per client to track usage independently.

API Key
Generate a key in your dashboard (starts with vbt_). Add it to your MCP client config. Revoke anytime.
OAuth 2.1
One-click connect in Claude Desktop. No API key to manage. Uses PKCE, dynamic client registration, and token rotation. The most secure option.

Frequently Asked Questions

Ready to connect

Your AI Assistant Is Ready to Speak

Connect Verbatik to your AI workflow in 30 seconds. No code. No subscription. Just add the endpoint and start talking.

Get Started — Connect Your AI ClientRead the Full Documentation
Free to set upNo subscription requiredPay only for what you generate