Your work. Your control.

Understand how access, shared AI colleagues and review work in Cedar—and what is still in development.

Choose what Cedar can access.

You choose which apps and sites Cedar can access. Available actions depend on the tools and permissions you enable. You can review or remove that access.

Share a colleague with clear boundaries.

An owner chooses who can work with a shared AI colleague and which learned rules to share. Sharing a colleague does not automatically share all of the owner’s conversations, files or connected apps.

The sharing controls distinguish work in the requester’s workspace, using their files and apps, from work that needs the owner’s workspace. What is returned follows the owner’s sharing and review choices.

Keep permission and judgment in the work.

AI colleagues follow their owners’ permissions. Some work can proceed under standing approval; other work needs approval for the specific task.

When review is needed, Cedar brings back the progress, relevant files and decision. Sharing controls can also require the owner to check a colleague’s contribution before it goes back to the requester.

Model processing and training.

Your data is not used to train AI models—by Cedar or its model providers. Cedar uses model APIs with no-training commitments.

A no-training commitment is separate from processing and retention. Those depend on the model provider and features used; this page does not claim that every task runs entirely on your Mac.

Read the privacy overview

About Parallel World.

Cedar for Mac

Some information is too sensitive to hand to anyone outside your company. Parallel World is how the most capable models work with it anyway.

Cedar’s safety model lives on your Mac. We trained it for this, and only this. Nothing of yours reaches the cloud without passing it, and it answers only to you.

The cloud sees the problem to be solved. What makes it yours stays on your Mac: your people, your clients, your numbers, your files, and anything you mark secret.

These are not promises of care. They are properties of how Cedar is built, and where it matters most they are proved: machine-checked in Lean.

After every task, you see exactly what the model saw.

In Cedar for Mac today, in English: your decisions and your documents.

See how it works

Ask us about your requirements.

For questions about access controls, model processing, deployment needs or a security concern, contact Cedar directly.

lei@cedarwork.ai