Running agents on a local model (Ollama)

Agents don't need cloud models.

Run them locally with Ollama:

• free, offline, private (data never leaves);
• can run on your edge devices (small models fit a Jetson).

How: most agent frameworks speak Ollama's unified API — just point the base URL at localhost.

Caveat: small models are weak at tool calling. → Get the loop working on a strong model first, then size down to optimize.

One line:

First make it work, then make it small.