Practical guides · 06
Run your terminal's AI on your own Mac with Ollama.
Set up a local model once, pick one that fits your Mac's memory, and know what to do when the engine won't start.
4 minute read
A cloud model is one option for terminal help. Another is a model that runs on your Mac. With a local provider such as Ollama, AITerm sends requests to an engine on the same computer instead of to a cloud provider, and there is no API key or per-request bill to manage. The trade-off is a one-time download and a model limited by your Mac's memory. This guide walks through the setup and the choices that matter.
1. Start from the setup door.
In AITerm's AI setup, choose Run AI on this Mac. What happens next depends on what is already installed:
- If Ollama is already running, Use Ollama reuses it with the models you have. If it has no chat model yet, Pick a model opens the Model Manager.
- If AITerm's own engine is installed but stopped, Start local AI starts it.
- If there is no engine at all, Download & set up offers to download Ollama and a model sized for your Mac. It shows the size first and waits for you to confirm. Nothing downloads at launch.
You can reach the same setup later from Settings ▸ AI.
2. Pick a model that fits your memory.
Press ⌘⇧M to open the Model Manager. It lists the models you have installed, lets you pull a new one by name (for example llama3 or mistral), delete ones you no longer use and switch the active model. Its Recommended for your Mac list is based on your Mac's memory, which is a sensible place to start. A larger model can give better answers but takes more disk space and memory, so try a recommended one before reaching for the biggest.
3. Check that it answers.
Look at the connection chip in the status bar. A green check means the connection has answered; an orange mark means something needs you. You can run a check yourself with Check in Settings ▸ AI. Then try a plain-English request, read the suggested command, and edit it before you run it, the same as with any model.
4. If the engine won't start.
The Local AI Engine panel in Settings ▸ AI shows the engine's status and offers Retry, reinstall or update. The handbook lists the common causes it names directly:
- Integrity check failed: use reinstall.
- Not enough disk space: free some space and retry.
- Offline during a download: reconnect and retry, or use a cloud key in the meantime.
If the selected connection can't start, AITerm turns AI off rather than quietly sending your request to a different provider. The model chip shows AI off; hover over it to read why.
When local makes sense.
A local model is a good fit when you are working near configuration files or logs you would rather keep on the machine, when you want to try AI help without setting up a provider account, or when you want help after the initial download without a network connection. A cloud model may still suit a harder question. You can keep both connections and switch in Settings ▸ AI. For how each path handles your session context, read What your AI terminal sends, and how to check it.
Try it in AITerm.
Local Ollama setup is part of the free core app, along with AI command suggestions, /fix, /explain, tabs, split panes, saved workspaces and supported bring-your-own-key connections. Runbooks and the autonomous agent are Pro. AITerm runs on Apple Silicon Macs with macOS 13 or later. Cloud provider charges are separate.
Keep learning.
Prepared with AI assistance from the current AITerm documentation. The steps follow the handbook; this is not a recorded app demonstration.