Open source · Local-first · macOS
The mailbox with your LLMs built in.
Ask your mail, sort it and draft replies with the models you already have. Fast, and stored on your computer.
- Claudeyour plan
- ChatGPTyour plan
- Ollamalocal
- + yours
What matters weighs more.
01 Ask
Type what you want.
An agent reads your mail and shows what fits, with a reason on every row.
3 of 212 fit · read on this Mac
02 Keep
Save it as a tab.
A saved prompt keeps reading new mail as it arrives.
has:attachment prompt:"invoices I still have to pay"
03 Agents
Your agents. Your plans.
It runs on the subscription you already pay for, or on a model on your Mac. With a plan, the mail the agent reads goes to that provider. With a local model, nothing leaves.
- Claude Codeyour Claude plan 87% left
- Codexyour ChatGPT plan signed in
- Ollamaon this Mac · mail never leaves it llama3.2
- API keysoon
- Your runtimeone small interface How →
04 Speed
Fast is the foundation.
The screen reads only a local database. The network is never in the way.
- < 500ms
- launch to inbox
- < 50ms
- open a thread
- < 16ms
- per search keystroke
Budgets, measured on mailboxes of 8,000, 100,000 and 500,000 messages.
05 Extend
One core. The rest plugs in.
Mail, storage and models are adapters behind small interfaces. Add one and the core stays as it is.
Write an adapter06 Next
Yet to come.
- Downloadssigned builds, updates
- Browser sign-inGmail, Microsoft 365
- API keysany provider
- Linux · Windowsbuilds and testers
- Yoursopen an issue ↗
07 Open source
Open. All of it.
Every decision is written down. Read it, run it, change it.
- AGPL-3.0 licence
- 120+ decision records
- 0 servers of ours
$ git clone https://github.com/milolabs/matters-app
$ cd matters-app
$ bun install
$ bun run dev:desktop:demo
✓ a made-up mailbox of 8,000 messages, no account needed
Get Matters.
Releases are on GitHub. Until the first one, you build it from source.