Workspaces and permissions
How organizations, workspaces, and roles control who and what can access your context.
Organizations and workspaces
An organization is the billing and membership boundary. Organizations have unique slugs, an invitation system, and roles: owner, admin, and member.
Organizations contain workspaces. A workspace is the unit of context: every file, folder, database, and connected source belongs to exactly one workspace. Workspaces have their own roles: admin, editor, and viewer.
Your current workspace ID appears in the workspace switcher in the sidebar. API tokens are bound to a single workspace, so agents never see across workspace boundaries.
Role-based access control
Web sessions are checked against both organization and workspace roles. Workspace admins manage members, connectors, and settings; editors change content; viewers read.
Agent access is separate: API tokens carry their own read, write, and delete permissions and an optional folder scope. A token can never do more than its permissions allow, regardless of who created it. See Tokens and scoping.
Plans and limits
Billing is per organization through Stripe. Owners and admins change plans from the Billing page, and on team plans the billed seat count is the seat allowance, so adding seats is a plan change rather than a separate setting.
Plan allowances cover seats, indexed content, asset storage, and connected sources; the pricing page lists the numbers per tier. A few limits show up while you work rather than on an invoice:
- Single asset size: 25 MB on Free, 1 GB on paid plans. This is checked during import preview, so an oversized attachment blocks the run before it starts.
- Migration archive size: 500 MB on Free, 5 GB on paid plans.
- Connected-source sync interval: once a day on Free, every six hours on Starter, hourly on Studio and Team, and every 15 minutes on Pro and Business. Paid plans also receive provider webhooks between scheduled runs.
- Revision history: 7 days on Free, 30 on Starter, 90 on Studio and Team, 180 on Pro, and 365 on Business.
Indexed content and asset storage are counted separately because they cost differently: indexed text expands into rows, embeddings, and vector indexes, while binary assets sit in object storage.