Connect VocaLingo to Claude, ChatGPT and Cursor via MCP

After you connect VocaLingo, your AI assistant can translate a voice message in the speaker's voice, summarize a YouTube video, voice a text in your own voice, dub a video or add subtitles. The agent spends tokens from your regular VocaLingo balance, and every result also appears in the app history: start a dub from your computer and pick it up on your phone.

Server details

Step 1. Open the AI agents screen

1. Open VocaLingo on iOS, Android or web.vocalingo.app and sign in.

2. Go to Settings → AI agents (MCP). The screen is called “Connect VocaLingo to your AI”.

3. Under “Choose your AI”, tap your assistant: Claude, ChatGPT, Cursor, VS Code, Claude Code or Other. The screen shows the steps for it.

Connect Claude

This works for claude.ai, Claude Desktop and the Claude mobile apps.

1. On the AI agents screen tap Open Claude. The “Add custom connector” window opens with the name and the address already filled in.

2. Tap Add, then Connect.

3. Claude opens the VocaLingo sign-in page. Sign in the same way as in the app: with a code sent to your email, with Google, or with a code from the app.

Manually: in Claude open Customize → Connectors → Add custom connector, paste https://nodejs.vocalingo.app/mcp and tap Connect. After that VocaLingo works in Claude on all your devices.

Connect ChatGPT

1. Copy the server address on the AI agents screen.

2. Open chatgpt.com/plugins in a browser, on your phone or computer.

3. Tap + → Add custom MCP server.

4. Enter the name VocaLingo, paste the address, choose OAuth sign-in, confirm “I understand and want to continue” and tap Create.

5. Sign in to VocaLingo on the page that opens.

Custom MCP servers are not available on every ChatGPT plan.

How to sign in from Claude or ChatGPT

The sign-in page accepts the same methods as the app: a code sent to your email, or Google. If you sign in to the app with Apple or don't remember how you signed in, tap Get a sign-in code on the AI agents screen. On the sign-in page choose “Sign in with a code from the VocaLingo app” and enter the 8-character code. The code works once and is valid for 10 minutes.

Connect Cursor or VS Code

1. Open web.vocalingo.app on the computer where Cursor or VS Code is installed and sign in.

2. Go to Settings → AI agents (MCP) and choose Cursor or VS Code.

3. Click Install in Cursor or Install in VS Code. The editor opens and offers to add the VocaLingo server: confirm.

One-click install works only on a computer. You can also tap Copy config and paste it into ~/.cursor/mcp.json for Cursor, or into mcp.json for VS Code (command “MCP: Open User Configuration”). The config already contains your token.

Connect Claude Code

1. On the AI agents screen choose Claude Code and tap Copy command.

2. Run it in the terminal. It looks like this:

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

Restart Claude Code. VocaLingo appears in the /mcp list.

Other MCP clients

Windsurf, Cline and any client that supports remote MCP servers connect with the address https://nodejs.vocalingo.app/mcp and the header Authorization: Bearer vlmcp_…. Choose Other on the AI agents screen to copy your token and a ready config. Plugins for Claude Code, Cursor, Codex and Gemini CLI with ready-made skills are in the vocalingo-mcp repository on GitHub.

VocaLingo MCP tools

The agent picks the tool from your request. Paid tools spend VocaLingo tokens, the in-app currency.

translate_video, caption_video and change_voice need a file. For a YouTube or TikTok link the agent first calls download_media and passes the downloaded file on. Prices of the paid tools are the same as in the app, see token prices for each tool.

How the agent asks about cost

• An operation that costs 50 tokens or more first returns the price, and the job starts only after you agree. The agent cannot confirm for you.

• You can ask the price in advance: “How much will it cost to dub this video?”. The agent gets an estimate without starting anything.

• Repeating the same request within 10 minutes does not charge twice: the agent gets the same job.

• If the balance runs out, the agent reports “insufficient balance”. Top up in the app, see how to buy tokens. There is no separate billing for MCP.

How to send audio, photos and videos to the agent

An uploaded file is kept for 48 hours, and the agent can reuse it without uploading again. Size limits: up to 150 MB for translate_audio, 50 MB for speech_to_text and save_voice, 100 MB for video tools and change_voice, 20 MB for images.

If you only need to read or translate the text on a photo, Claude and ChatGPT do that themselves for free. translate_image is for when you want the picture itself with the text redrawn.

Long jobs

Long tools wait about 25 seconds and return the result if it is ready. Otherwise the agent gets a jobId and checks it with check_job. Dubbing can take up to 30 minutes. If you closed the chat during a job, ask the agent to call check_job without a jobId: it lists recent jobs. Links to result files are valid for 24 hours; the results themselves stay in the app history (without Premium, the last 3 items in each tool).

Access, token and how to turn it off

Anyone with your token can spend tokens from your balance, so keep it secret. The Access and tokens section of the AI agents screen has two actions:

• Create a new token: the old token stops working within a minute. Replace it in the agents connected by token.

• Disable MCP access: all agents lose access within a minute, including the Claude and ChatGPT connectors. To connect again, tap Enable MCP access, and a new token is issued.

Clone only your own voice or a voice you have explicit permission to use. The same rule applies in the app.

Frequently asked questions