Skip to main content
Glama
matheusslg

MCP Audio RAG Server

by matheusslg
OverviewSchemaRelated ServersScoreDiscussions
TypeScript
Remote

MCP Audio RAG Server

Transform your audio files into a searchable knowledge base using AI. Ask Claude questions about your meetings, podcasts, lectures, or any audio content.

What is this?

This is an MCP (Model Context Protocol) server that lets you:

  1. Transcribe any audio file using Google's Gemini AI

  2. Store the transcriptions in a searchable database

  3. Search through all your audio content using natural language

Once set up, you can simply ask Claude things like:

  • "What did they discuss about the budget in my meeting recording?"

  • "Find mentions of machine learning in my podcast collection"

  • "What were the key points from yesterday's lecture?"

How It Works

┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │ Audio File │ ──▶ │ Gemini │ ──▶ │ Chunking │ ──▶ │ Supabase │ │ (.mp3, etc) │ │ Transcribe │ │ + Embedding │ │ (pgvector) │ └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘ │ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │ │ Claude │ ◀── │ Results │ ◀── │ Search │ ◀──────────┘ │ Response │ │ + Snippets │ │ Query │ └─────────────┘ └─────────────┘ └─────────────┘

Quick Start

Prerequisites

Step 1: Clone & Install

git clone https://github.com/matheusslg/mcp-audio-rag.git cd mcp-audio-rag npm install

Step 2: Set Up Supabase Database

  1. Create a new project at supabase.com

  2. Go to SQL Editor in your dashboard

  3. Paste and run the contents of supabase/schema.sql

Step 3: Get Your API Keys

Supabase (Settings → API):

  • Copy Project URL → SUPABASE_URL

  • Copy service_role key → SUPABASE_SERVICE_KEY

Google AI Studio:

Step 4: Configure

cp .env.example .env

Edit .env:

GEMINI_API_KEY=your-key-here SUPABASE_URL=https://your-project.supabase.co SUPABASE_SERVICE_KEY=your-service-role-key

Step 5: Add to Claude

For Claude Code CLI (~/.claude.json):

{ "mcpServers": { "audio-rag": { "command": "npx", "args": ["tsx", "/full/path/to/mcp-audio-rag/src/server.ts"], "env": { "GEMINI_API_KEY": "your-key", "SUPABASE_URL": "https://your-project.supabase.co", "SUPABASE_SERVICE_KEY": "your-service-role-key" } } } }

For Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json on Mac):

Same config as above.

Usage

Transcribe Audio

Just tell Claude to transcribe a file:

Transcribe /path/to/meeting.mp3

Want to use a specific model? Just ask:

Transcribe /path/to/lecture.m4a using gemini-2.5-pro

Search Your Audio

Ask natural questions:

What did they say about the project timeline? Search for mentions of "budget" in my recordings Find discussions about AI in my podcasts

Manage Your Library

List all my transcribed audio files Delete the recording from last week Get the full transcript of meeting.mp3 Summarize the podcast episode

Available Models

Model

Best For

gemini-2.5-flash

Default - Fast & accurate, great balance

gemini-2.5-flash-lite

Fastest, cheapest - good for bulk processing

gemini-2.5-pro

Best quality - complex audio, multiple speakers

gemini-3-pro-preview

Newest - cutting edge capabilities

gemini-2.0-flash

Reliable - previous generation

gemini-2.0-flash-lite

Fast - previous generation

Supported Audio Formats

.mp3 .mp4 .m4a .wav .webm .mpeg .mpga

Available Tools

Tool

Description

ingest_audio

Transcribe and store an audio file

search_transcripts

Search through your audio using natural language

list_transcripts

List all transcribed audio files

get_full_transcript

Get the complete transcript of a file

summarize_audio

Generate an AI summary of a transcript

delete_transcript

Remove a transcribed file from the database

Troubleshooting

Problem

Solution

"No relevant segments found"

Try rephrasing your search, or check if audio was ingested

"Missing environment variable"

Check your .env file or Claude config has all 3 keys

Supabase errors

Make sure you're using service_role key, not anon key

Slow transcription

Use gemini-2.5-flash-lite for faster processing

Support This Project

If this project saved you time or helped you out, consider buying me a coffee!

License

MIT - Use it however you want!

This server cannot be installed

-
security - not tested
F
license - not found
-
quality - not tested

How are these scores calculated?

Resources

New MCP Servers

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/matheusslg/mcp-audio-rag'

If you have feedback or need assistance with the MCP directory API, please join our Discord server