Ollama’s v0.32.0-rc0 adds an agent UI to the CLI — the single change listed in the release notes.
What shipped:
- Agent UI in CLI — PR #17017, the only entry in the changelog
- Release candidate tag —
rc0means first candidate, not stable - No model updates — no new model weights or quantization changes listed
Ollama runs LLMs locally (Llama, Qwen, Gemma, Mistral) through a single ollama run command. The CLI has been a straightforward chat interface. An agent UI in the command-line tool points to tool-use and multi-step interaction — capabilities developers previously got through third-party wrappers or separate frameworks.
For developers using Ollama as a local coding backend, this matters. Ollama pairs with Continue in VS Code, routes through Cursor’s custom model config, and powers offline workflows. An agent UI in the core CLI could handle simple agentic tasks without a separate frontend — though the release notes don’t detail what the UI includes.
NOTE
This is rc0 — the first release candidate. The agent UI is likely rough. Don’t run it against production workflows or critical local inference setups.
The pre-release is live on GitHub. Stable users should wait for the final v0.32.0 tag.