Email replies from a local model: Ollama or LM Studio.
If you already run a model in Ollama or LM Studio, you can use it to draft email replies. paisty is a menu-bar app for Apple Silicon Macs that reads the message in front of you and sends the draft request to the model you point it at. Point it at your local model and the request stays on the computer you configured.
Setup in short
- Start the model in Ollama or LM Studio, on the same Mac or on a computer you already reach on your network.
- Open paisty setup and choose the local model option. Nothing is connected until you press a button.
- Optional: add a voice file at ~/.config/paiste/voice.md for tone.
- Open a message, click the paisty mark, read the draft, send it yourself.
If your server speaks the OpenAI-compatible API, you can also enter its URL as an OpenAI-compatible endpoint.
Picking a model
Email replies are short, so a small instruction-tuned model can be enough for routine mail. Larger models usually follow the thread better and make fewer mistakes, but they need more memory and take longer. We do not publish benchmark numbers; try two or three models on your own mail and keep the one whose drafts you edit least. You can switch back to a cloud model in setup any time.
What stays local, and what does not
- The window text, any notes you connected, and the draft go to the model on the computer you configured, not to HailPorts.
- The app still checks your license key with paisty.com. That check sends the key, not the window text.
- If your model runs on another computer, the request travels over your network to it.
Full detail: privacy.
Honest tradeoffs
- Local drafts can be weaker than large cloud models, especially on long threads.
- Speed depends on your Mac and the model size.
- You maintain the model server. paisty does not install Ollama or LM Studio for you.
Price
A local model has no per-token bill. paisty: Free 7-day trial, one per email. Then $6/month, $60/year, $99 lifetime, or $49 for a Founding Member lifetime key through December 31, 2026. Checkout is Stripe. Details: pricing.
More: your own AI, including local models and privacy-first AI email drafts.
FAQ
- Does paisty work with Ollama? Yes. Choose a local model in setup and point paisty at the Ollama model you already run, on that Mac or another computer you reach.
- Does paisty work with LM Studio? Yes. Use a model already running in LM Studio, or enter its OpenAI-compatible server URL.
- Does any email text leave my Mac with a local model? The draft request goes to the local model on the computer you configured. The app separately checks your license key with paisty.com, which sends the key, not the window text.
- Which local model is best for email? It depends on your Mac and your mail. Try a few and keep the one whose drafts you edit least. We do not publish benchmarks.
Related
- An AI email reply app for Mac with a one-time price
- Reply to client emails faster: AI drafts for agencies on a Mac
- Recruiter email reply templates, plus AI drafts on a Mac
- All answers
Pricing · About · Support · AI email drafts · Compare apps · Your own AI · vs Superhuman · vs Shortwave · vs ChatGPT desktop · vs Raycast AI · AI in any text box · Best Mac reply tool · All answers · Privacy · Terms