Use Claude in Apple Mail
MailTwin is a menubar assistant for Apple Mail. You type a few keywords and it writes the reply, in your own style once you have indexed your sent mail. Its provider setting decides which AI writes. This page covers Claude, from Anthropic, used with your own API key. For models that run on your Mac instead, see the Ollama and LM Studio guides, or the full provider guide.
What you need
- A Mac running macOS 13 or later, with MailTwin installed and Apple Mail set up as usual.
- An Anthropic API key from console.anthropic.com. API use is billed by Anthropic separately from any Claude.ai subscription; MailTwin's in-app note mentions a $5 minimum to start.
Setup
- Create a key. Sign in at console.anthropic.com, add billing, and create an API key. Claude keys start with
sk-ant-; MailTwin warns you if a pasted key does not. - Open the provider settings. Right-click the MailTwin menubar icon, choose Settings, then AI and Providers, and open the Claude card.
- Save and test. Paste the key and click Save & Test. MailTwin stores the key in your macOS Keychain and sends a short test request; the card shows the reply, or the reason it failed.
- Make it active. Click Make active. MailTwin never switches provider on its own.
- Optional: choose a model. The default is
claude-opus-4-8. To use another Claude model, open Advanced on the card and type its model id.
Using it
- Select a message in Apple Mail and press ⌘⇧M to open the MailTwin panel. Type a few keywords, press Return to generate, and Return again to insert. Claude replies stream in as they are written.
- Or stay in Mail's own reply window: type your keywords, press the MailTwin button and choose Expand keywords.
- Settings → Privacy lists each request MailTwin made, so you can confirm Claude answered and see what it received.
What is sent, and to whom
The request goes directly from your Mac to Anthropic, signed with your key. There is no MailTwin server in the AI path, and MailTwin has no AI proxy. Anthropic receives the text needed for the action you asked for: for a reply, the email you are answering and your keywords; for a reply in your style, also a few short examples from your sent mail, which MailTwin keeps in a local index on your Mac. MailTwin's provider note says Anthropic does not train on API requests by default and that requests go to Anthropic's US servers; Anthropic's own terms are the authority.
If names in your mail should not reach a cloud provider, turn on Settings →
Privacy → Mask names before cloud requests. It replaces names,
email addresses and phone numbers with placeholders such as [PERSON_1]
on your Mac and puts them back in the reply. It can miss unusual names, so it
reduces what is sent rather than making it anonymous. For mail that should not
leave the Mac at all, use an on-device provider.
Cost
Anthropic bills you directly. Its published price for the default model is $5 per million input tokens and $25 per million output tokens (checked 13 October 2026; prices change). Settings → Usage shows MailTwin's running estimate, which is approximate; Anthropic's bill is authoritative. For a worked comparison across providers, see the AI email reply cost calculator.
Frequently asked
Can I use Claude to write replies in Apple Mail?
Yes. Add your Anthropic API key in Settings → AI → Providers, make Claude active, and generate replies from the MailTwin panel or Mail's reply window.
Does the email go through a MailTwin server?
No. The request goes from your Mac to Anthropic with your key.
Does my Claude.ai subscription work?
No. MailTwin uses the Anthropic API, which needs an API key and is billed separately from a Claude.ai plan.
Trial
The 30-day trial covers every provider. MailTwin itself is a one-time purchase; Claude usage is billed by Anthropic. Get MailTwin.