Text to speech in 149 languages: MP3 links from any assistant. Free without an account.
One address for any MCP client. No sign-in needed on this server.
3 read4 other
Grouped by tool name.
Odel charges nothing per call. The publisher may charge for its own service.
Text to speech inside Claude, ChatGPT, Cursor, VS Code, Codex, Windsurf, Gemini CLI and any other MCP client. One remote server, nothing to install:
https://mcp.freetts.org/mcp
Ask your assistant to read something aloud, and it calls FreeTTS and hands you an MP3 link.
| Tool | What it does | Needs |
|---|---|---|
text_to_speech | Text in, MP3 link out. 2,253 voices in 149 languages. | Nothing. Free without an account. |
list_voices | Voices for a language or search word, with ids. | Nothing |
suggest_voice | Three voices for a language and a use, with reasons. | Nothing |
check_usage | Plan and characters left. | Nothing |
dialogue_to_speech | Name: line script, a voice per speaker, one file. | FreeTTS PRO |
script_to_tracks | Script mode: exact (pause 2), (beep), (ding), TRACK: files, CUE: voice, scenes. | FreeTTS PRO |
transcribe_audio | A public audio URL in, text out. | FreeTTS PRO |
Free without a key: standard voices, 2,000 characters a call and a day per connection, 10 requests a minute, a short spoken "FreeTTS" tag at the end, files kept one hour.
With a FreeTTS key (free account or PRO): the account's plan. PRO: HD and Signature voices, no tag, 10,000 characters a call, 1,000,000 a month, WAV, files kept 30 days, and the dialogue, Script mode and transcription tools. Keys: freetts.org/dashboard (API keys).
Claude Code
claude mcp add --transport http freetts https://mcp.freetts.org/mcp
With a key: add --header "Authorization: Bearer ft_live_...", or run /mcp inside a session to connect with the sign-in flow.
Claude (web, desktop, mobile): Settings, Connectors, Add custom connector, paste https://mcp.freetts.org/mcp. Free tools work at once; when a tool needs your account Claude shows Connect.
Cursor: Install in Cursor, or in ~/.cursor/mcp.json:
{ "mcpServers": { "freetts": { "url": "https://mcp.freetts.org/mcp" } } }
VS Code: Install in VS Code, or in .vscode/mcp.json:
{ "servers": { "freetts": { "type": "http", "url": "https://mcp.freetts.org/mcp" } } }
Codex CLI and ChatGPT desktop
codex mcp add freetts --url https://mcp.freetts.org/mcp
ChatGPT (web): Settings, Apps (Developer Mode), add https://mcp.freetts.org/mcp.
Windsurf, ~/.codeium/windsurf/mcp_config.json:
{ "mcpServers": { "freetts": { "serverUrl": "https://mcp.freetts.org/mcp" } } }
Any other client: use https://mcp.freetts.org/mcp as the server URL (Streamable HTTP). To use your account without the sign-in flow, send your key as Authorization: Bearer ft_live_... or x-api-key: ft_live_....
Full guide with copy buttons: freetts.org/developers/mcp.
The server is an OAuth 2.1 authorization server (PKCE, dynamic client registration, client ID metadata documents, RFC 9728 and 8414 metadata). Free tools need no sign-in. When a client calls a tool that needs your account, the server answers 401 and the client shows its Connect flow; the page asks for a FreeTTS API key from your dashboard and binds the connection to it. Delete the key in the dashboard to disconnect every assistant that used it.
The text you send is sent to FreeTTS and its speech providers exactly as from the website, under the FreeTTS privacy policy. Calls without a key are metered by the caller's IP address. The server logs tool name, plan, character count, client name and outcome, never the text.
The same server also runs on your own machine, for clients that start local servers or when you would rather keep a key in a local config. It talks to the public FreeTTS API with your own key. Get one at freetts.org/dashboard (API keys; a free account has one).
FREETTS_API_KEY=ft_live_... npx -y @freetts/mcp
Claude Desktop, Cursor, Cline, Windsurf and similar clients:
{
"mcpServers": {
"freetts": {
"command": "npx",
"args": ["-y", "@freetts/mcp"],
"env": { "FREETTS_API_KEY": "ft_live_..." }
}
}
}
Without a key it still starts: list_voices, suggest_voice, the prompts and the resources work, and making audio asks for a key. The hosted server at https://mcp.freetts.org/mcp needs no key at all.
From source: npm install && npm run build && node dist/stdio.js. Node.js 20 or later.
The hosted server itself (dist/index.js) runs next to the FreeTTS API on FreeTTS's own machines and is not meant to be self-hosted; the code is here so you can read exactly what it does.
License: MIT.