A Streamable-HTTP MCP Server that uses memvid to encode text data into videos that can be quickly looked up with semantic search.

Supported Actions:

• add_chunks: Adds chunks to the memory video. Note: each time you add chunks, it resets the memory.mp4. Unsure if there is a way to incrementally add.
• search: queries for the top-matching chunks. Returns 5 by default, but can be changed with top_k param.

Set up your environment:

python3.11 -m venv my_env
. ./my_env/bin/activate
pip install -r requirements.txt

Run the server:

python server.py

With a custom port:

PORT=3002 python server.py

You can connect a client to your MCP Server once it's running. Configure per the client's configuration. There is the mcp-config.json that has an example configuration that looks like this:

{
  "mcpServers": {
    "memvid": {
      "type": "streamable-http",
      "url": "http://localhost:3000"
    }
  }
}

• Obviously the modelcontextprotocol and Anthropic teams for the MCP Specification. https://modelcontextprotocol.io/introduction
• HeyFerrante for enabling and sponsoring this project.