In development. Mac first, then Windows and Linux.
The right local AI for your computer.
A private AI app for your files and tools. Sevra will choose and update the model for you.
Built on Slotstream. Our open-source engine runs a 105 GB model on a Mac with 48 GB of memory, loading parts from SSD as needed. It is native Swift on Apple’s MLX and Metal, tuned as one system for that model.
See the code and measurementsPick up where you left off.
Come back to a project, ask what changed and turn your notes into next steps. Sevra is being built to work with the files and tools you give it access to.
Your spaceChapters
Website launch
Research notes
Weekend plans
Your journal
Website launch
The launch moved to Friday, and the signup flow still needs a review. Here are the changes I’d make to your task list.
From your Planning notes and Launch checklist
Task list
Website launch
Move the due date from Thursday to Friday
Review the signup flow
Add to your tasks before launch
2 changes ready for your review
Keep the context that matters
Your conversation will have chapters you can search and revisit, with a journal for notes you want to keep. You’ll be able to see what Sevra remembers, correct it or stop it from using a detail.
Make your own tools
You’ll be able to ask Sevra for a small app, such as a project tracker, or save a repeatable task as a skill. Your data will stay in your files when an app changes.
We take care of the models.
Chosen for your computer
Sevra will match the model and settings to your computer. We’ll test answer quality, speed and tool use on hardware like yours, and share the results and tradeoffs.
Tuned for everyday work
We’ll tune the app around a few carefully chosen models, from longer conversations to reliable tool use.
Kept up to date
We’ll compare new models before offering updates. You’ll see what changes, decide when to update and know which earlier version you can return to.
Runs on your computer.
After the model download, you’ll be able to work offline. Your conversations and personal files will stay on your computer. You won’t need an account to use Sevra locally.
Sevra is being built natively for each platform, from the app down to the engine, so it uses your computer’s memory and graphics chip directly. On Mac, that engine is Slotstream, and the app will call it in process.
You decide what it can do.
You’ll choose which files and tools Sevra can use, and what it can change. You’ll also be able to see which model is running and limit how much of your computer it uses.
Before you join
Can I download Sevra yet?
Not yet. Join the waitlist for development updates and early-access announcements. We haven’t announced a release date or price.
Will it run on my computer?
We’re starting with Apple Silicon Macs. Windows and Linux are planned next, including AMD and NVIDIA support. We’ll publish the exact memory, storage and device requirements before each release.
Do I need to choose or set up a model?
Sevra will recommend a tested model for your computer and guide you through the download. You won’t need to compare models or tune settings yourself.
Get development updates and early-access news.
Join by email. You can unsubscribe at any time.