Use FireAI’s AI on an Intel Mac with Ollama

FireAI works on Macs with an Intel processor. Its AI (Ask FireAI for any order, AI summaries in Investigate, explanations of AI agent alerts) needs one more free app there: Ollama. Ollama runs Gemma 4 E2B, the same model FireAI uses on Apple silicon, on your Mac’s processor. FireAI talks to Ollama only inside this Mac (address 127.0.0.1), so what you type stays on the Mac.

Without Ollama, every other FireAI feature works, and Ask FireAI still understands simple orders such as “block Teams” or “go in paranoid mode”.

What you need

• macOS 14 Sonoma or later.
• About 5 GB of free disk space for the model.
• 8 GB of memory at least; 16 GB is better. The model uses about 5.5 GB of memory while it answers, and Ollama frees it after 10 minutes without a question.

Set it up

Open FireAI, go to Home and look under Ask FireAI. The “Use FireAI’s AI on this Intel Mac” panel walks you through three steps and ticks each one off when it’s done.

1. Install Ollama. Click “Download Ollama”: your browser opens ollama.com/download. Download Ollama for macOS, open the file and drag Ollama to Applications. FireAI never installs anything by itself.
2. Open Ollama. Click “Open Ollama”, or open it from Applications; its icon appears in the menu bar. The first time, Ollama may offer to install its command-line tool: you can accept.
3. Download the model. Click “Download the model” (Gemma 4 E2B, about 4.6 GB). Ollama downloads it once, keeps it on your Mac and shows the progress. You can also run this command in Terminal instead:

ollama pull gemma4:e2b

When the three steps are done, the panel disappears and Ask FireAI shows “AI ready · runs only on this Mac”. If a step isn’t detected, click “Check again”.

FireAI’s firewall and Ollama

• FireAI never blocks or asks about connections inside your Mac, so it can always talk to Ollama.
• When you click “Download the model”, FireAI adds two rules that let Ollama reach its model library (ollama.ai). You’ll find them in Rules, in the Assistant group, and you can delete them after the download.
• If you download the model from Terminal instead, FireAI may ask about Ollama’s connection like any other app’s. Click Allow.
• The kill switch blocks every connection, including this download. Turn it off while the model downloads.

What it can’t do

• It doesn’t install Ollama or the model for you: you click each step, and FireAI checks it.
• It doesn’t answer with AI while Ollama is closed. If you quit Ollama, the setup panel comes back; open Ollama again.
• It never sends your orders to an online AI instead. If Ollama is missing, or hasn’t answered after 90 seconds, FireAI uses its built-in understanding, which handles simple orders only.
• It can’t run Gemma 4 on an old Ollama. If yours is too old, FireAI asks you to update it from ollama.com.
• It isn’t meant for Macs with less than 8 GB of memory.

Good to know

• The first question after Ollama starts waits while the model loads; the next ones are quicker.
• Installed Ollama’s command-line version with Homebrew? Start it with “ollama serve” in Terminal, then click “Check again”.
• On a Mac with Apple silicon, FireAI runs the AI itself and doesn’t need Ollama. If FireAI was opened with Rosetta, turn off “Open using Rosetta” in FireAI’s Get Info window.