MCP server
Deepgram
Send audio to Deepgram's speech-to-text engine and get a transcript back inside Claude.
Published by the company that makes the product.
What it can access
Sends whichever audio file you provide to Deepgram's API using an account key you supply; no access to anything else on your machine.
Security note
Published by Deepgram under an MIT licence, from Deepgram's own GitHub organisation. Unlike Whisper on your own computer, this sends audio to Deepgram's servers. Do not use it to transcribe an interview or testimony recording: use the offline whisper-local entry in the tool catalogue for that.
Before you start
This one needs a terminal (a command-line window) and Node.js already on your computer. If a terminal is unfamiliar to you, install a Skill from this directory instead, or ask a technical colleague to do these steps with you.
To check whether Node.js is installed: open the Terminal app (PowerShell on Windows), type node -v, and press Enter. A version number means it is installed. “command not found” means you have to install it from its own official website first.
How to install
- Follow the “Installation” section of the project’s own instructions at github.com/deepgram/mcp.
- You will be asked to add a short block of text to Claude Desktop’s configuration file and restart the app.
- After restarting, open Claude Desktop’s Settings and confirm “Deepgram MCP Server” is listed as connected.
How to remove it
- Open the configuration file from step 2, delete the block you added for “Deepgram MCP Server”, save it, and restart Claude Desktop.
- Delete or revoke any API key or account you created for this on the vendor’s own website. Otherwise a stray copy of the configuration file is still a working key.