Every Workspace Finally Has a Real AI Policy Screen
Workspace admins can now set a real spending cap and approval threshold for their AI agents from Settings — a page that used to just say 'coming soon.'
For a while, the workspace configuration page in Settings said exactly one thing: "Workspace configuration coming soon." That's no longer true. The first real thing to land there is something admins have been asking for directly — a working AI policy screen that actually controls what your agents are allowed to do.
What you can set
From your workspace settings, an admin can now define a spending cap per agent run, a list of forbidden capabilities the agent is never allowed to use in this workspace, and an approval threshold that decides when an action needs a human to sign off before it happens versus when the agent can just proceed. None of this is new machinery bolted on top — the underlying policy system has existed since earlier this year, enforced by the runtime every time an agent runs. What was missing, until now, was a screen where an admin could actually see and change it.
One rule worth knowing
Every field in this policy can only get stricter, never looser, relative to what's already in place. You can lower a spending cap, add to the forbidden list, or tighten when approval is required — but nothing here can be used to quietly widen what an agent is allowed to do. That's a deliberate constraint, not a current limitation: a policy screen that could loosen restrictions as easily as tighten them wouldn't really be a policy screen.
Why this matters
Agent capability has been moving faster than agent governance across the industry, and this platform is no exception — this is a real instance of governance catching up to what agents can already do. If your workspace runs AI agents and nobody has explicitly set spending limits or approval rules, this is now the place to do it.
This is live today for every workspace on the platform.