CEX

Project / About

Intelligence with a history.

Bring your intelligence. Give it a life that carries forward: experience to draw on, skills to reuse, and work that grows with you.

The idea

Give your intelligence a life

CEX is a free, local AGI harness. You bring the model; CEX gives it memory, tools, skills, goals, and a place to keep working. Useful context and lessons from earlier work can become part of what the system brings to the next task.

The aim is continuity that grows with you: less rebuilding context, more drawing on what came before, and a shared history that remains when you change models.

What we mean by AGI harness

The harness is the persistent system around a model, not a claim that every model becomes AGI. Learning here means carrying forward saved experience, feedback, and reusable skills; it does not mean CEX automatically retrains the underlying model.

The distinction

Build on the work you have already done

Remember a preference. Reuse a method. Return to a goal with its progress intact. CEX brings these parts together so intelligence can work with a history, with permissions you can inspect and control.

Model-independentRun a local model or connect a provider with credentials you control.
Local firstRuntime state, workspaces, and fresh memory stores stay on the machine by default.
InspectableTool calls, policies, goals, and runtime state are designed to be visible in the product.
Experience that carries forwardSaved lessons and reusable skills give future work something to draw on.
Founder-stage software

Built before it was polished

CEX began as a solo project on consumer hardware. The architecture is real; packaging, onboarding, documentation, and platform coverage are still being hardened.

The Linux beta is public first because it is the build that exists today. Windows and macOS will not be presented as available until their native packages are built and verified.

Early does not mean imaginary.

The product is usable, but the project is still learning where the runtime holds up and where it gets in the way. Specific, reproducible feedback is part of the build process.