Admin Controls (Workspace Agents) — Definition
Admin Controls are the governance layer in a ChatGPT Business or Enterprise workspace that defines who can build agents, which connectors they can use, and what's logged.
Longer definition
Workspace Agents inherit the governance posture of their host ChatGPT tenant, plus agent-specific controls: per-connector approvals scoped to user groups, explicit "builder" roles (not every seat), audit log retention (30 days on Business, configurable on Enterprise), per-agent credit caps, and offboarding flows when an agent owner leaves the company.
In practice
A well-governed Workspace Agents deployment has: (1) a specific group of authorized agent builders, (2) connectors approved per-group with least-privilege scopes, (3) audit log retention matched to compliance requirements, (4) per-agent caps on credits and on write actions, (5) a named owner for every agent with an offboarding procedure.
Common mistakes
- Leaving agent-building open to every seat — leads to sprawl and ungoverned agents
- Default 30-day audit retention when compliance needs 90+ days
- No offboarding procedure — when the agent owner leaves, nobody notices the agent is now orphaned
See also
Want this built for your team?
20-min intro call. $1,000 per agent, ~1 week to production.