Integration · Local LLMs & Ollama

Private, on-device memory for local models

Running Ollama or another local model? Kiomon gives it the same durable memory that frontier models enjoy, fully on-device without cloud round-trips.

Connect

  • Start Kiomon's MCP server locally: 
    npx kiomon-mcp --token <your-kiomon-token> --url <your-kiomon-url>
  • Point your MCP client (Claude Desktop, Cursor, or any client that fronts Ollama) at the server.
  • Your local model now has full read/write access to the memory graph through the standard toolset.

Why local models need memory most

  • Local models have smaller context windows, making persistent memory essential for long tasks.
  • Everything stays in your control: the MCP server runs on your machine; the token scopes exactly what the model can touch.
  • Zero training on your data by default. With local models, your raw data never leaves your hardware.

Ready

Create your free workspace or read the protocol reference. Building something custom? The REST API speaks the same memory layer.