MCP Server for Rewind AI Transcripts

A Model Context Protocol (MCP) server that provides tools to query and search through Rewind AI screen recording transcripts.

Features

• 🔍 Search transcripts by date and time range
• 📅 List available dates with transcript data
• 📊 Get summaries of transcript activity for any date
• 🔎 Full-text search within transcript content
• ⏰ Time-based filtering to focus on specific periods

Installation

Using uv (recommended)

# Clone the repository
git clone https://github.com/yourusername/mcp-server-rewind.git
cd mcp-server-rewind

# Install with uv
uv pip install -e .

Using pip

# Clone the repository
git clone https://github.com/yourusername/mcp-server-rewind.git
cd mcp-server-rewind

# Install with pip
pip install -e .

Configuration

Claude Desktop Integration

Add the following to your Claude Desktop configuration file:

MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "rewind": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/mcp-server-rewind",
        "run",
        "mcp-server-rewind"
      ],
      "env": {
        "REWIND_BASE_PATH": "/Users/yourusername/Library/Application Support/com.memoryvault.MemoryVault/chunks",
        "REWIND_TIMEZONE": "Europe/Berlin"
      }
    }
  }
}

Environment Variables

• REWIND_BASE_PATH: Base path to Rewind chunks directory (default: ~/Library/Application Support/com.memoryvault.MemoryVault/chunks)
• REWIND_TIMEZONE: Timezone for timestamp interpretation (optional, e.g., "Europe/Berlin")

Usage

Once configured, the following tools are available in Claude:

1. Search Transcripts

Search for screen recording transcripts by date with optional time range and text query.

search_transcripts:
  date: "2025-05-31"
  start_time: "09:30"  # Optional: HH:MM format
  end_time: "10:30"    # Optional: HH:MM format
  query: "neuraflow"   # Optional: text to search for

2. List Transcript Dates

List all dates that have transcript data, optionally filtered by year and/or month.

list_transcript_dates:
  year: 2025  # Optional
  month: 5    # Optional (1-12)

3. Get Transcript Summary

Get a summary of transcript activity for a specific date.

get_transcript_summary:
  date: "2025-05-31"

Example Queries

Find all screen activity for a specific time period:

"Show me what I was working on between 9:30 and 10:30 on May 31st"

Search for specific content:

"Find all mentions of 'neuraflow' in my screen recordings from May 31st"

Get an overview of recording activity:

"What dates do I have screen recordings for in May 2025?"

Summarize a day's activity:

"Give me a summary of my screen recording activity on May 29th"

How It Works

The server reads transcript files from Rewind's local storage, which are organized by date in the following structure:

chunks/
├── 202505/          # YYYYMM
│   ├── 29/          # DD
│   │   ├── video_file_transcript_20250529_093756_to_20250529_094410.txt
│   │   └── ...
│   └── 31/
│       └── ...
└── ...

Each transcript file contains the OCR'd text from the corresponding video segment, allowing you to search through your screen content as if it were a document.

Development

Running Tests

uv run pytest

Code Quality

# Type checking
uv run pyright

# Linting
uv run ruff check .

# Formatting
uv run ruff format .

License

MIT License - see LICENSE file for details.

Privacy Note

This tool only accesses local Rewind transcript files on your machine. No data is sent to external services. All processing happens locally on your computer.