CloudBolt Software announced the addition of AI agent governance capabilities to the CloudBolt Cloud Management Platform (CMP). The capabilities build on the Model Context Protocol support CloudBolt announced in May and the natural language interface that came with a CMP redesign in August. They include the following:
- Every agent acts as a named user. OAuth support maps each AI agent to an identity in CMP. The agent gets that user’s groups, environments, rights, and limits, with no shared service account holding broad credentials behind it.
- Agents order from the governed catalog. Agents see only the blueprints that administrators have published, the same catalog people use in the CMP portal.
- Administrators control access action by action. Each CMP action carries its own AI setting. Teams can open routine work to agents while keeping high-impact actions available only to people in the portal.
- Enforcement happens at execution, for any agent. CMP checks every agent action against the user’s permissions and policies, regardless of how the request was worded or which AI model sent it.
- Every agent session can be traced. An agent dashboard in CMP records each session, including the tools an agent called, the orders it placed, and the user it acted for.
Agent governance capabilities are currently available to all CloudBolt CMP customers.
For more information, go to cloudbolt.io.