Discover ThinkOS
The operating layer for agents you own.
ThinkOS is a desktop-first workspace where your AI agents live with you — with your memory, your context, and your rules. It runs local-first on your machine, and nothing about how you work is trapped inside someone else’s product silo.
Public beta. Free to join. Feedback welcome.
How it works
Your memory, your workflows, any model.
Most AI tools are powerful in isolation. ThinkOS is the layer around them — the place where memory, workflows, and permissions become portable and useful across everything you do.
A memory vault you can open
Your Work Memory lives on your device as plain Markdown — readable in any editor, friendly to Git and backups, and portable by default. Every update keeps its provenance, so you can see where a memory came from.
Workflows agents can carry
Repeatable jobs like a daily operator brief, meeting capture, and project briefs run as inspectable workflows. You can watch every step, and durable memory writes wait for your approval.
Any model, one harness
ThinkOS is model-agnostic. It routes each job to the right model — local or hosted — through one consistent harness, so switching models never means rebuilding how you work.
Connected to your real work
Bring in files, email, calendar, and docs so agents work from your actual context — the people, projects, decisions, and meetings that matter — instead of one-off prompts.
Trust you can verify
The Trust Center shows exactly what the app is doing: telemetry is opt-in, audit events record agent activity, and export, revoke, and purge are always one action away.
Desktop first, everywhere next
ThinkOS runs on macOS and Windows with a mobile companion app that pairs to your desktop. A bundled agent CLI means other agents and tools can operate ThinkOS too — no UI required.
The public beta
Help shape ThinkOS from the start.
The public beta is free and open to anyone ready to try ThinkOS on macOS, Windows, or Linux. It is still an early release: rough edges are expected, and bug reports, ideas, and candid feedback will directly shape what we build next.
Public beta. Free to join. Feedback welcome.