Your workspace

A workspace is more than a tenant boundary. It holds an agent that works for you, a file tree that agent writes into, the environments your evals run in, and the registry that says where their images came from.

• Workspace agents — the agent that runs inside Everdict, and how it wakes itself
• What the agent knows — three layers of context, and time as a coordinate
• The workspace filesystem — one file tree per workspace, with attributed revisions
• Environments — repo · prompt · browser · os-use, and which one a task deserves
• Secrets — every credential, by name, encrypted at rest
• Image registry — publishing images, and the provenance that makes two scorecards comparable

These four connect: an agent writes its output to the filesystem, an eval runs in an environment, and that environment is an image whose reference decides whether last week's number means anything.

• Browser profiles — a captured session, so a case is about the task rather than the login

For the boundary itself — roles, secrets, isolation — see Workspace.