Connect a local model (Ollama) to atbridge.ai

With Ollama, the model runs on your machine, so the atbridge AI Chat processes your email content locally and never sends it to any AI provider. This is the most private way to use the atbridge AI Chat — and there’s no API key to manage.

Note

Install atbridge.ai and the Thunderbird add-on first — see Install atbridge.ai. New to the in-Thunderbird side? Start with Set up the atbridge extension.

Prefer to watch? Here's the full walkthrough:

Connect Ollama

1. Install Ollama and pull a model

1. Install Ollama from ollama.com — it runs on macOS, Windows, and Linux.

2. Pull a model in a terminal. A solid all-round default is Qwen:

   ollama pull qwen3

   Other good choices are llama3.1 and mistral.

3. Ollama runs locally at http://localhost:11434 — leave it running.

Want sharper answers? Go bigger — if your machine can handle it.

Larger models reason noticeably better for email tasks. If you have the RAM (and ideally a GPU), pull a bigger Qwen variant — e.g. ollama pull qwen3:14b or qwen3:32b. It depends on your machine: bigger models need more memory and run slower on modest hardware, so start with qwen3 and size up only if it feels comfortable.

Tip

Because the model runs on your machine with Ollama, your email content stays on your machine — nothing goes to a cloud AI provider. There’s also nothing to pay per use: it’s your own hardware.

2. Connect it in atbridge.ai

1. In Thunderbird, open the atbridge Settings — the gear, or Tools → Add-ons and Themes → atbridge.ai → Options (see Set up the atbridge extension). Go to the AI providers tab.

2. Pick Ollama from the Provider dropdown.

3. There’s no API key needed — atbridge.ai discovers your installed models automatically.

4. Click Connect. The status turns to a green ✓.

3. Pick your local model and chat

Open the atbridge AI Chat (the @b button in Thunderbird), pick your local model from the model dropdown, and ask away:

“Summarise the unread thread from Maya and draft a reply.”

With Ollama selected, that request is handled by the model on your machine.

Tip

Want the end-to-end privacy walkthrough? See the deeper recipe Run a fully local AI (Ollama).

Related

• Connect your AI — overview
• Use the atbridge AI Chat
• Run a fully local AI (Ollama) recipe