Which Ollama models work best?

Agents rely on tool calling, so choose a model that supports tools in Ollama. Smaller models chat fine but may struggle with multi-step tool use.

Ollama runs on another machine — can I still use it?

Yes. Add the provider with that machine's address as the API base, for example http://192.168.1.20:11434/v1.

Integration · Models

Point FastClaw at Ollama to run AI agents on local, open-weight models — no API key, no data leaving your machine.

The ollama preset targets http://localhost:11434/v1 through Ollama's OpenAI-compatible endpoint. Nothing to sign up for.

FastClaw stores data in a local SQLite file and Ollama runs the model, so prompts and memory stay on your hardware.

Run private agents on Ollama and give others a hosted model — each agent picks its own provider.

Run a private personal agent on your laptop with no cloud dependency — the developer setup.
Prototype agents for free before switching to a hosted model like OpenAI or DeepSeek.
Keep sensitive agents on local models while others use cloud providers.

ollama pull qwen3

In Console → Models, click Add provider and choose Ollama. The API base defaults to http://localhost:11434/v1; no API key is needed.
Or from the CLI:

fastclaw agents init local \
  --provider ollama \
  --model ollama/qwen3

If FastClaw itself runs in Docker, localhost refers to the container — use the host's address for the API base instead.

Run FastClaw agents on OpenAI models — set a deployment-wide default, override per agent, and expose your agents through an OpenAI-compatible API.

Run FastClaw agents on Claude through Anthropic's native Messages API — with prompt caching, extended thinking and image input preserved.

Run FastClaw agents on DeepSeek — a built-in preset, thinking-mode support and per-agent model overrides for low-cost, capable agents.

Connect FastClaw to OpenRouter and give every agent access to models from many vendors with a single API key.

Start free in FastClaw Cloud, or self-host the single binary in a minute.