Deepgram MCP server
Deepgram in any MCP client, through Memtro: Deepgram does not publish this server. Memtro's Deepgram connector reads Deepgram through its API with the credentials you connect, and one endpoint hands those tools, every other system you connect and Memtro's memory tools to Claude Code, Cursor, Claude Desktop or anything else that speaks Streamable HTTP, as you.
What the model can do
Deepgram speech AI: the projects the key can see, each project's request log (when, which endpoint, status, never what was said), usage by hours, requests, tokens and characters, billing in dollars grouped by line item, tag, key or deployment, balances, the public speech and voice models, and the voice agent configurations. When asked, can transcribe an audio file by URL (billed per audio hour), create, relabel or delete a voice agent configuration, and rename a project.
Reads
deepgram_projectsDeepgram projects this API key can see: id and name.deepgram_requestsA Deepgram project's request log, one page at a time: when each request was made, the endpoint path and which features it used, the status code…deepgram_usageA Deepgram project's usage between two dates: audio hours, total hours, agent hours, LLM tokens in and out…deepgram_billingWhat a Deepgram project was billed between two dates, in US dollars, optionally grouped by line item (e.g. streaming::nova-3), tag, key…deepgram_balancesA Deepgram project's outstanding balances: amount, units (e.g. USD) and purchase reference.deepgram_modelsDeepgram's public models: speech-to-text (name, canonical name, architecture, languages, version…deepgram_agent_configsA Deepgram project's saved voice agent configurations: id, labels (metadata), and when each was created and updated.
Acts, when you ask
These change something in Deepgram. The model is told to use them only when you ask for exactly that, and each one is granted by name: per team on a shared connection, by the owner on a personal one.
deepgram_transcribe_urlTranscribe an audio or video file at an https URL with Deepgram and return the transcript (cut to 20,000 characters) with any analyses asked for.deepgram_create_agent_configSave a reusable Deepgram voice agent configuration in a project, from the agent block of a Settings message given as JSON.deepgram_update_agent_labelsReplace the labels (metadata) on a saved Deepgram voice agent configuration.deepgram_delete_agent_configDelete a saved Deepgram voice agent configuration.deepgram_rename_projectRename a Deepgram project.
Connect it
First connect Deepgram on Dashboard → Connections, then add Memtro to your client once. The Deepgram tools appear alongside the tools of everything else you have connected; nothing is configured per connector.
Claude Code
The Memtro plugin registers the server and a skill for when to use it; the tools are then available in every session.
claude plugin marketplace add memtro-org/claude-memtro
claude plugin install memtro@memtro --config api_key=sk-memtro-...Cursor
Add the server to .cursor/mcp.json in the project, or to the global file under ~/.cursor/.
{
"mcpServers": {
"memtro": {
"url": "https://platform.memtro.com/api/mcp",
"headers": {
"Authorization": "Bearer sk-memtro-..."
}
}
}
}Claude Desktop
Create a personal API key on Dashboard → Profile → API keys. In Claude Desktop add an MCP server of type HTTP with URL https://platform.memtro.com/api/mcp and the header Authorization: Bearer sk-memtro-…
OpenCode
A remote server in opencode.json.
{
"mcp": {
"memtro": {
"type": "remote",
"url": "https://platform.memtro.com/api/mcp",
"headers": {
"Authorization": "Bearer sk-memtro-..."
}
}
}
}Pi
One command: pi mcp add memtro --url https://platform.memtro.com/api/mcp --bearer-token-env-var MEMTRO_API_KEY. Tools arrive as mcp__memtro__*. The pi-memtro extension registers it for you, alongside Memtro as the model.
Any other client
Add an MCP server of type HTTP (Streamable HTTP, stateless) at https://platform.memtro.com/api/mcp with the header Authorization: Bearer sk-memtro-.... Codex, Zed, Windsurf, Cline and the SDKs all take a remote server this way.
Who may use it
The key identifies the person, and the server offers that person the tools they may use everywhere else in Memtro. A shared Deepgram connection is used by the teams an administrator granted it to, tool by tool, on the Permissions page; a personal connection is used by its owner alone, and its tools that act stay off until the owner allows them. See Teams and permissions and Connectors.
Ask it like this
- › How many audio hours did we use this month?
- › What were we billed last month, by model?
- › Which requests failed yesterday?
- › What is our outstanding balance?
- › Transcribe the call recording at this URL
Questions
- Is this an official Deepgram MCP server?
- No. It is Memtro's MCP server with Memtro's Deepgram connector behind it: the connector reads Deepgram through its API with the credentials you connect on Dashboard → Connections, and Deepgram publishes nothing here. The same endpoint carries every other system you connect and Memtro's memory tools.
- What does it cost?
- Nothing on top of Memtro. One price per organisation by its number of users, from $50 a month for up to 10 users after a 7-day free trial, with every connector, the MCP server, jobs and memory included. You bring your own model keys.
- Can it change anything in Deepgram?
- Only when you ask, and only what it has been granted: transcribe an audio file at a URL (billed per audio hour plus features); create a voice agent configuration; change a voice agent configuration's labels; delete a voice agent configuration (anything using its id stops working); rename a project. Every other tool reads, searches or lists. An administrator grants each acting tool by name per team, and a personal connection acts only after its owner allows it.
- Which clients work?
- Any MCP client that speaks Streamable HTTP: Claude Code, Cursor, Claude Desktop, OpenCode, Codex, Zed, Windsurf, Cline and the Anthropic and OpenAI SDKs. There is no local process to run; the server is remote and authenticated with your key.