Roadmap
LCOA is in beta. This page says what works today, what is being finished, and what is only planned. A plan is not a promise.
Works today, in the beta
- Chat, writing and maths with models that run on your PC, and web search when you ask for it.
- Voice: you speak and it speaks back, with listening and voices that run on your PC.
- Safe action: you choose how much LCOA may touch, it shows what it will change before it changes it, and most changes can be undone.
- Tidying files, a task list, reminders and your calendar and mail (Google is optional; mail is only ever saved as a draft).
- Memory you can see: what you ask it to remember is kept on your PC, and you can ask it to forget.
- LCOA Code, a project agent that proposes each edit and waits for your yes.
- Connecting other programs’ tools through MCP. Off until you add a program yourself; every call shows you the exact request first.
Being finished before and just after launch
- Windows code signing, so the installer stops showing a publisher warning.
- A smoother first run: a starter model picked for your PC.
- A real Mac test. The Mac version is a first release that has not been tried on many real Macs.
- Published speed and quality results for the models that ship in LCOA, with the method.
- Hosting for sign-in that does not depend on one computer being switched on, and a support address on lcoa.ai.
Planned, not promised
- An optional model in the cloud that you connect with your own key, off by default and clearly marked when it is on.
- Longer conversations: a larger context window for the models that can take it.
- Linux.
Something you need that isn’t here? Say so through support.