Connect your AI (in Thunderbird)

The @bridge AI Chat inside Thunderbird runs on the AI of your choice. You bring your own API key once; from then on the chat uses that provider to work your mail and calendar.

Note

Your key stays yours. It’s encrypted on this device and used to talk to the provider you chose, directly. atbridge.ai never routes your key — or your mail — through its own servers. Prefer to keep everything local? Use Ollama, where the model runs on your machine too.

The flow is the same for every provider

1. Get an API key from your provider (links below).

2. In Thunderbird, open @bridge Settings → AI providers (see Set up the @bridge extension).

3. Pick the provider, paste the key, click Connect → green ✓.

4. Open the @bridge AI Chat, choose a model, and start asking.

You can connect several providers and switch between them (and their models) right in the chat — your keys all stay local.

Pick your provider

OpenAI GPT-4o and others. Key from platform.openai.com.

Anthropic (Claude) Claude Sonnet / Opus / Haiku.

Google Gemini Gemini Flash / Pro. Free tier available.

Local model (Ollama) No key, no cloud — fully private.

CLI subscription (Claude Code, Antigravity, Codex) Already pay for a coding-agent CLI? Use it — no key, nothing metered. Paid plans.

Other providers Mistral, Groq, DeepSeek, xAI, OpenRouter, NVIDIA, OpenAI-compatible.

Where to get each key

Each provider’s key comes from its own site. Direct links:

Provider	Get your API key	Full guide
OpenAI	OpenAI platform	Connect OpenAI
Anthropic (Claude)	Claude developer platform	Connect Anthropic
Google Gemini	Google AI Studio	Connect Gemini
Mistral	Mistral admin console	Other providers
Groq	Groq console	Other providers
DeepSeek	DeepSeek platform	Other providers
xAI (Grok)	xAI console → API Keys	Other providers
OpenRouter	OpenRouter keys	Other providers
NVIDIA NIM	build.nvidia.com — free, no card (not private)	Other providers
Ollama (local)	No key — install Ollama	Connect Ollama
Claude Code / Antigravity / Codex CLI	No key — sign in to the CLI you already have (paid plan)	Connect a CLI

What it costs

You pay two things: your atbridge.ai plan (there’s a Free tier; Personal and Pro are paid subscriptions) and your AI provider’s per-token usage, billed directly by the provider you chose. atbridge.ai puts no markup on that provider usage — and because it sends lean, text-first requests, each task uses far fewer tokens than tools that dump full HTML. See Why @bridge is so fast and cost-effective.

Don't want to pay per use?

Run a model locally with Ollama — it’s free (your own hardware) and private (your email content stays on your machine). A great default, and the answer whenever a cloud provider’s pricing gives you pause.

Related

• Use the @bridge AI Chat
• Set up the @bridge extension
• Plans & pricing