Integration · Local LLMs & Ollama

A local model with durable memory

Running Ollama or another local model? Point your stack at Kiomon's hosted API and it gains durable memory across sessions. The model runs on your hardware; the memory layer is Kiomon's hosted service.

Connect

  • Kiomon is reached over its hosted OpenAI-compatible API — the same surface any OpenAI client uses: 
    base_url = "https://api.kiomon.com/v1"
api_key  = "<your-kiomon-token>"
  • Point your local stack at that base URL — an Ollama-fronted app, LangChain, n8n, or any OpenAI-compatible client. Completions still come from your local model; only memory reads and writes go to Kiomon.
  • Kiomon's MCP toolset is hosted too, at https://mcp.kiomon.com/mcp, for remote-capable MCP clients such as Cursor. Ollama is not itself an MCP client, so use the API path above for local-model stacks.

Why local models need memory most

  • Local models have smaller context windows, making persistent memory essential for long tasks.
  • Scoped by key: your API key decides exactly which workspaces the model can read and write.
  • No training on your data, ever. Your model runs on your hardware; memories you save are stored in your Kiomon workspace, and prompts sent to Kiomon's grounded-answer endpoint are processed by Kiomon to return citations.

Ready

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