AI assistants (MCP)
The official JSF Labs MCP server lets AI assistants like Claude, Cursor and other MCP clients use JSF Labs directly. Ask in plain language: "read this paragraph in an Urdu female voice and save it to my Desktop". The assistant picks a voice, generates the audio and saves the file for you.
It uses your normal API key, so the usual credits and rate limits apply.
Tools
| Tool | What it does |
|---|---|
list_default_voices | Browse the default voices by language, gender or name |
list_my_voices | List your cloned voices |
text_to_speech | Generate speech, and optionally save the file locally |
clone_voice | Clone a voice from a local WAV or MP3 file |
delete_voice | Delete one of your cloned voices (asks for confirmation) |
- 1Get your API key
Copy your
jsf_key from your JSF Labs dashboard. You also need Node.js 18 or newer. - 2Add the server to your assistant
{ "mcpServers": { "jsf-labs": { "command": "npx", "args": ["-y", "@jsflabs/mcp-server"], "env": { "JSF_API_KEY": "jsf_your_api_key" } } } }In Claude Desktop, open Settings → Developer → Edit Config. In Cursor, edit
~/.cursor/mcp.json. - 3Restart and ask
Restart the app, then try one of these:
- "Show me Urdu female voices on JSF Labs"
- "Say 'Welcome to JSF Labs' with Aaliya and save it to my Desktop"
- "Clone my voice from ~/Desktop/me.wav and call it Jawad"
The key sits in your local config file and is only sent to JSF Labs. Don't share that config file or commit it to a repository.
Some assistants stop waiting for a tool after a few minutes. For very long scripts, split the text into parts or call Text to speech directly.
