Skip to main content
Glama
evans-sam

fireflies-mcp

by evans-sam
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript
Remote

Fireflies MCP Server & CLI

MCP server and CLI for the Fireflies.ai API. Retrieve transcripts, search meetings, and generate summaries.

Forked from Props-Labs/fireflies-mcp.

Quick Start

# Install globally
npm install -g fireflies-mcp
# or
bun install -g fireflies-mcp

# Set your API key
export FIREFLIES_API_KEY=your_key

# Use the CLI
fireflies-mcp transcripts --limit 5
fireflies-mcp search "standup" --json
fireflies-mcp details <transcript-id>
fireflies-mcp summary <transcript-id>

Related MCP server: Fathom AI MCP

Commands

fireflies-mcp                              Start MCP server (stdio, default)
fireflies-mcp serve                        Start MCP server (stdio, explicit)
fireflies-mcp serve --http [--port 3000]   Start MCP HTTP server
fireflies-mcp transcripts [options]        List recent transcripts
fireflies-mcp details <id> [options]       Get transcript details
fireflies-mcp search <query> [options]     Search transcripts by title
fireflies-mcp summary <id> [options]       Generate transcript summary
fireflies-mcp --help                       Show help
fireflies-mcp --version                    Show version

CLI Options

Option

Applies to

Description

--json

all CLI commands

Output raw JSON instead of formatted text

--limit <n>

transcripts, search

Maximum results (default: 20)

--from <date>

transcripts, search

Start date filter (YYYY-MM-DD)

--to <date>

transcripts, search

End date filter (YYYY-MM-DD)

--format <fmt>

summary

bullet_points (default) or paragraph

MCP Tools

When running as an MCP server, exposes these tools:

  1. fireflies_get_transcripts — List transcripts with optional date/limit filters

  2. fireflies_get_transcript_details — Full transcript with speakers, text, and summary

  3. fireflies_search_transcripts — Search by title keyword

  4. fireflies_generate_summary — Summary in bullet points or paragraph format

Installation

Option 1: npx (no install)

npx fireflies-mcp transcripts --limit 5

Option 2: Global install

npm install -g fireflies-mcp
# or
bun install -g github:evans-sam/fireflies-mcp

Option 3: From source

git clone https://github.com/evans-sam/fireflies-mcp.git
cd fireflies-mcp
bun install

MCP Server Configuration

Claude Code (~/.claude.json)

Via npx (Node.js):

{
  "mcpServers": {
    "fireflies": {
      "command": "npx",
      "args": ["fireflies-mcp"],
      "env": { "FIREFLIES_API_KEY": "<YOUR_API_KEY>" }
    }
  }
}

Via Bun (from source):

{
  "mcpServers": {
    "fireflies": {
      "command": "bun",
      "args": ["run", "/absolute/path/to/fireflies-mcp/src/main.ts"],
      "env": { "FIREFLIES_API_KEY": "<YOUR_API_KEY>" }
    }
  }
}

Via Docker (HTTP):

{
  "mcpServers": {
    "fireflies": {
      "type": "http",
      "url": "http://localhost:3000/mcp"
    }
  }
}

Claude Desktop

Same JSON structure — add to claude_desktop_config.json.

Docker

# docker-compose (recommended)
FIREFLIES_API_KEY=your_key docker compose up -d

# or directly
docker build -t fireflies-mcp .
docker run -d -p 127.0.0.1:3000:3000 -e FIREFLIES_API_KEY=your_key --restart unless-stopped fireflies-mcp

# verify
curl http://localhost:3000/health

Setup

Fireflies API Key

Get your key at fireflies.ai/dashboard/settings/api.

Development

bun install          # Install deps
bun run start        # Start MCP server (stdio)
bun test             # Run tests
bun run lint         # Lint
bun run format       # Format (auto-fix)
bun run build        # Build for Node.js

Publishing

Publishing is automated via GitHub Actions. To release:

# Bump version in package.json, then:
git tag v0.2.0
git push origin v0.2.0

This triggers CI → test → build → npm publish → GitHub release.

Requires NPM_TOKEN secret in the repository settings.

Environment Variables

Variable

Required

Default

Description

FIREFLIES_API_KEY

Yes

—

Your Fireflies.ai API key

License

MIT. See Props-Labs/fireflies-mcp for the original.

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Resources

Related MCP Connectors

Related MCP Servers