LocalInbox can write summaries, suggested replies, drafts and answers with a model you run on your own Mac in Ollama, LM Studio or another OpenAI-compatible server. Choose it in Settings › Intelligence. LocalInbox accepts only an address on this Mac, so your mail stays on it, with no API key and no AI cloud. It works on Macs without Apple Intelligence too.
Needs a model server running on your Mac, such as Ollama or LM Studio, with a model downloaded. LocalInbox doesn't download models itself.
What your model writes
- A one-line summary of recent conversations in the message list, and key points beside the conversation.
- Three suggested replies under a message, answering its newest part.
- Draft with AI, from a sentence describing the email you want.
- Rewrite: shorter, friendlier, more formal, or with the grammar fixed.
- Ask your mailbox: answers from your own mail, with numbered sources.
Set it up in a minute
Three steps. The Ollama and LM Studio guides go through each one, with what to do if Test Connection doesn't connect.
- Install Ollama or LM Studio and download a model.
- In LocalInbox, open Settings › Intelligence and set Written by to “A model server on this Mac”.
- Click Ollama or LM Studio to fill in the address, choose a model, and click Test Connection.
Why it stays private
- Only addresses on this Mac are accepted: 127.0.0.1, ::1 or localhost. Anything else is refused before a request is made.
- Requests skip any system proxy and never follow a redirect, so they can't be carried off the Mac.
- The address and model are kept on this Mac, not synced to iCloud.
- Cloud AI isn't an option: LocalInbox takes no OpenAI, Anthropic or other API keys.
What stays the same
Sorting mail into Primary, Updates, Promotions and your categories, To Reply and phishing checks don't depend on the model server. Rules decide them, with Apple's model where it's available.
Questions
Which model should I use?
Any chat model your server can run. Smaller models answer faster; larger ones write better. Try one, then use Test Connection, which shows which model answered and how long it took.
Does it work without Apple Intelligence?
Yes. A model server is the way to get summaries, replies, drafts and Ask on a Mac without Apple Intelligence, including Macs on macOS 15.
What if the server isn't running?
LocalInbox treats the AI features as unavailable until it's back, and checks again within a minute. Everything else keeps working.
Can a model server send my mail to the cloud?
LocalInbox only ever talks to the server on your Mac, but some servers can hand requests on to cloud models, such as Ollama's cloud models. Choose a model that runs on your Mac and your mail stays on it.
Can I use a model server on another computer?
No. LocalInbox refuses any address that isn't on this Mac, so your mail never crosses the network to reach a model.