Connect VocaLingo to AI agents (MCP server)

Give Claude Code, Cursor or any AI agent access to VocaLingo: voice cloning, audio and video translation, subtitles and summaries via MCP.

Written By Umakhan Magomedov

Last updated 1 day ago

VocaLingo has an MCP server: connect it to your AI agent (Claude Code, Cursor, Claude Desktop and others) and the agent can save your voice, voice texts, translate audio and video, add subtitles and summarize videos. Payment comes from your regular VocaLingo balance.

Step 1. Get your token

Open the VocaLingo app (web, iOS or Android) and sign in with your email. Guest accounts cannot create a token.
Go to Settings, then API / MCP Token.
Pick your client (Cursor, Claude Code, Claude Desktop or Other) and tap Copy config for your agent: you get a ready JSON snippet in the right format for that client.

⚠️ Anyone with this token can spend tokens from your balance. Keep it secret. If it leaks, tap Regenerate token: the old one stops working within a minute. You can also fully turn off agent access with Disable MCP access on the same page.

Step 2. Connect your agent

Server URL: https://nodejs.vocalingo.app/mcp. It is a remote server, nothing to install.

Claude Code

claude mcp add --transport http vocalingo https://nodejs.vocalingo.app/mcp --header "Authorization: Bearer vlmcp_YOUR_TOKEN"

Cursor: add to ~/.cursor/mcp.json:

{  "mcpServers": {    "vocalingo": {      "url": "https://nodejs.vocalingo.app/mcp",      "headers": { "Authorization": "Bearer vlmcp_YOUR_TOKEN" }    }  }}

Claude Desktop: the connector settings have no field for the Authorization header yet, so connect through the mcp-remote bridge (requires Node.js). Open Settings, Developer, Edit Config (file claude_desktop_config.json) and add:

{  "mcpServers": {    "vocalingo": {      "command": "npx",      "args": ["-y", "mcp-remote", "https://nodejs.vocalingo.app/mcp", "--header", "Authorization:${AUTH_HEADER}"],      "env": { "AUTH_HEADER": "Bearer vlmcp_YOUR_TOKEN" }    }  }}

Note: Authorization:${AUTH_HEADER} is written without a space on purpose. After saving, fully restart Claude Desktop (Quit, not just close the window). To let Claude upload your chat attachments, also add sb.vocalingo.app in Settings, Capabilities, Code execution, Additional allowed domains.

Claude.ai in the browser is not supported yet: it cannot pass the Authorization header to custom connectors. Use Claude Desktop instead.

Other clients (Windsurf, VS Code, OpenClaw, Hermes and any MCP client with remote server support): use the same URL and header.

What the agent can do

Tool	What it does	Price
save_voice	Saves your voice from an audio sample	~150 tokens
text_to_speech	Voices any text with your saved voice, OpenAI or ElevenLabs voices	~10 tokens per 1 000 characters
speech_to_text	Transcribes audio to text	~1 token per 2 min
translate_audio	Translates audio, optionally keeping the original voice	depends on length
summarize_media	Transcript and summary of a video or audio, including YouTube, TikTok and Instagram links	~1 token per 2.5 min
translate_video	Dubs a video into another language with voice cloning and lip-sync	~220 tokens per minute
caption_video	Burns styled subtitles into a video, can translate them	~1-2 tokens per minute

Free tools: get_upload_url, check_upload, list_voices, delete_voice, check_job, list_jobs, get_balance, get_history, get_transactions.

Everything the agent does also appears in your app history: start a video translation from your computer, get the result on your phone.

💡 Before expensive operations the agent can request a cost estimate (dryRun) and will show it to you before spending tokens. You can also ask it directly: "how much will this cost?"

How to send files to the agent

Links: YouTube, TikTok, Instagram and direct file links work as is, just paste them in the chat.
Local files: agents with a terminal (Claude Code, Cursor) upload them automatically via a secure one-time link.
No terminal: the agent sends you a browser upload page, drop the file there and continue the conversation.
Telegram bot assistants: voice messages forwarded to your bot are passed by direct link, zero extra steps. This is the most seamless setup.

Acceptable use

Voice cloning through the API follows the same policy as the app: clone only your own voice or a voice you have explicit consent to use. Uploaded files are deleted automatically after 48 hours; results stay in your app history.

Frequently asked questions

Claude Desktop has no field for the token. Where do I put it?

That is expected: the custom connector UI currently supports only URL and OAuth. Use the mcp-remote config from Step 2 in claude_desktop_config.json and fully restart Claude Desktop.

The agent says "insufficient balance". What do I do?

Top up the balance in the VocaLingo app (Profile, Balance). The MCP server uses the same balance as the app, there is no separate API billing.

I closed the chat while a long job was running. Is the result lost?

No. Ask the agent to call list_jobs: it will find the job and fetch the result. The result is also saved in your app history.

Can I have several tokens?

One token per account. Regenerating replaces the old one everywhere.

Why does my agent fail to upload a file in Claude Desktop?

Add sb.vocalingo.app to Settings, Capabilities, Code execution, Additional allowed domains. Without it Claude cannot upload attachments to VocaLingo.

A result link from the agent has expired. Is the file gone?

No. Links are valid for 24 hours, but the result itself is stored in your app history. Ask the agent to call check_job with the same jobId to get a fresh link.

Still need help? Share an idea