MCP Server

Use the Deepgram CLI as an MCP server for AI coding tools.

The dg CLI includes a built-in MCP (Model Context Protocol) server that gives AI coding tools direct access to Deepgram APIs.

Start the MCP Server

$dg mcp

Starts a server using stdio transport (default).

HTTP/SSE Transport

For HTTP-based MCP clients:

$dg mcp --transport sse --port 8000

Configuration

Claude Code

Add to your project’s .claude/mcp.json:

1{
2 "mcpServers": {
3 "deepgram": {
4 "type": "stdio",
5 "command": "dg",
6 "args": ["mcp"]
7 }
8 }
9}

Or add globally:

$claude mcp add deepgram --scope user --command dg --args mcp

Cursor

Add to .cursor/mcp.json:

1{
2 "mcpServers": {
3 "deepgram": {
4 "type": "stdio",
5 "command": "dg",
6 "args": ["mcp"]
7 }
8 }
9}

Windsurf

Add to .windsurf/mcp.json:

1{
2 "mcpServers": {
3 "deepgram": {
4 "type": "stdio",
5 "command": "dg",
6 "args": ["mcp"]
7 }
8 }
9}

Available Tools

The MCP server exposes these capabilities:

  • transcribe_audio — Transcribe audio from URL or file path
  • synthesize_speech — Generate speech from text
  • analyze_text — Sentiment, topics, summarization
  • list_models — List available Deepgram models
  • get_usage — Check API usage

Authentication

The MCP server uses your stored credentials (dg login) automatically.

For CI environments:

$DEEPGRAM_API_KEY=sk_xxx dg mcp