RewindMCP Server

You set up and run an MCP STDIO server that exposes RewindDB capabilities to GenAI models. This server lets you fetch transcripts, OCR content, and activity data from RewindDB so AI tools can analyze and summarize your data with natural language queries.

How to use

Install and run the MCP STDIO server locally, then connect your MCP-compatible client to it. You will typically start the server, then send requests to retrieve transcripts, OCR content, screen data, or activity statistics. Use the provided environment variables to point the server at your RewindDB database and secure access.

How to install

# prerequisites
python3 -m pip install --upgrade pip
# create and activate a virtual environment (optional but recommended)
python3 -m venv .venv
source .venv/bin/activate

# install the Rewind MCP package in editable mode (from source)
pip install -e .

# install MCP dependencies for development and running the STDIO server
pip install mcp>=0.1.0

MCP server configuration

Configure and run the MCP STDIO server to expose RewindDB tools to your AI clients. The server reads database connection details from environment variables and supports standard MCP tooling.

{
  "mcpServers": {
    "rewinddb": {
      "command": "python",
      "args": ["/path/to/mcp_stdio.py"],
      "env": {
        "REWIND_DB_PATH": "/path/to/your/rewind.db",
        "REWIND_DB_PASSWORD": "your_password"
      }
    }
  }
}

Run the MCP STDIO server

Start the server using the command shown in the configuration snippet. You can also pass a custom environment file if you keep sensitive values in a separate file.

python mcp_stdio.py

Connecting a client

Use an MCP-compatible client to communicate with the server. The client will discover available tools, such as getting transcripts, searching transcripts, and retrieving OCR and activity data. Pass the appropriate parameters to each tool to retrieve exactly the data you need.

Tools exposed by the server

The server provides a collection of tools that let you access different data sources within RewindDB. These tools enable you to fetch transcripts, search within transcripts, retrieve screen OCR content, obtain activity statistics, and drill into transcript details by ID.

Security and environment considerations

Keep your .env values secure and avoid publishing database paths or passwords. Use a dedicated environment file or secret manager for the REWIND_DB_PATH and REWIND_DB_PASSWORD variables, and restrict access to the MCP server endpoint to trusted clients.

Troubleshooting

If the server fails to start or respond, verify that MCP dependencies are installed, confirm the database configuration in the environment, and check logs for detailed error messages.

Available tools

get_transcripts_relative

Fetch audio transcripts for a relative time window, useful for analysis and model training.

get_transcripts_absolute

Fetch complete transcripts for a specific absolute time window, ideal for meeting summaries.

search_transcripts

Search transcripts for keywords or phrases, returning matches with context.

search_screen_ocr

Search OCR text content from screen captures for keywords across a time range.

get_screen_ocr_relative

Retrieve all screen OCR content for a relative time period, optionally filtered by application.

get_screen_ocr_absolute

Retrieve all screen OCR content for a specific time window, with optional application filter.

get_ocr_applications_relative

Discover applications with OCR data in a relative time period, with activity metrics.

get_ocr_applications_absolute

Discover applications with OCR data in a specific time window, with activity metrics.

get_activity_stats

Get activity statistics covering audio, screen captures, and application usage for a time period.

get_transcript_by_id

Retrieve the full transcript content for a specific audio ID after obtaining a preview.