MCP Server

The VocalLab MCP server lets AI tools — Claude Desktop, Claude Code, Cursor, VS Code and others — generate speech with your account through the Model Context Protocol.

Watch the full walkthrough — covers the API and MCP

Requires a Pro plan or higher. No API key needed up front — the setup creates one for you.

Quickest setup (one click)

On the Developer page, open Use in AI tools (MCP). Pick your client, then press the single button — it generates an API key for you and wires it up in one step. There's nothing to copy or paste.

Client	Button	What happens
Cursor	Add to Cursor	Opens Cursor and installs the VocalLab server directly.
VS Code	Add to VS Code	Opens VS Code and installs the server directly.
Claude Code	Copy install command	Copies a `claude mcp add …` command — paste it into your terminal or a Claude Code chat.
Claude Desktop	Download config file	Saves a ready-to-use `claude_desktop_config.json` to drop into your Claude config folder.

Your new key is shown once in the banner at the top of the section. Copy it somewhere safe before you leave the page — for security we can't show it again.

Prefer to set it up by hand? Every client also has Copy and Download buttons on the config preview, so you can paste the snippet wherever you like.

Using VocalLab Studio

In VocalLab Studio, open API & MCP in the sidebar and switch to the MCP tab — same idea: generate a key, pick your client, then connect. It also supports Windsurf and any other MCP-compatible client via a generic config.

Tools

get_balance — your points balance
list_voices — voices you can use
list_models — the speech models (VocalLab Studio / Studio Flash / Pro / Lite)
generate_speech — text → speech, returns an audio URL. Accepts an optional model (v-studio, v-flash, v-pro, or v-lite)
get_speech — look up a generation by id: status + download/stream links
get_captions — SRT subtitles for a generation (phrase lines or word-by-word)
delete_speech — permanently delete a generation you own
design_voice — describe a voice in words and get up to 3 candidate previews to listen to
save_designed_voice — keep a preview as a permanent voice you can generate with
clone_voice — clone a voice from a short recording (passed as base64)
list_voice_languages — languages and recording limits for cloning and design
delete_voice — permanently delete one of your cloned or designed voices

Once connected, ask your assistant to "list VocalLab voices", "generate speech with VocalLab", "design me a warm British narrator voice", or "generate that with the VocalLab Studio model." See Voice Models for what each model does.

Hosted server

Every one-click option points at the hosted server — there's nothing to install or run:

URL: https://mcp.vocallab.ai/mcp
Header: Authorization: Bearer vl_live_...

Manual config (Claude Desktop & others)

This is exactly what the Download config file button produces. Add it to your client's MCP config:

{
  "mcpServers": {
    "vocallab": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.vocallab.ai/mcp", "--header", "Authorization: Bearer vl_live_..."]
    }
  }
}

Claude Desktop's config file lives at:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

If the file already exists, merge the vocallab entry under mcpServers instead of replacing it, then restart Claude Desktop.