Governing AI agents
How do you control what an AI agent is allowed to do, on whose behalf, at the moment it acts?
Agent identity, delegated authority, MCP and tool access, runtime policy and decisions that weigh history across calls. This is where most teams start: governing the agents they already run.
Identity & access
Who the agent acts for, how authority narrows at each hop, and how far its reach goes.
Tenant isolation for AI agents and RAG: why a tenant_id filter isn't enough
No other customer's text was shown. Another customer still changed the decision. How tenant boundaries leak through retrieval, caches, models, tools and logs, and how to hold them end to end.
Runtime policy
Rules enforced before and after every step, deciding with history, not just the call in front of them.
Governance by design for AI applications: why trust can't be bolted on
Every component was secured. The order still went to the wrong legal entity. Why AI governance added after the prototype fails, and what it means to build it into how the application runs.
Visual policy rules for AI governance: no hidden scripts behind the diagram
The rule on screen required a second approver. A hidden script quietly turned it off. How to build visual policy rules whose facts, outcomes, conditions and tests are the same thing that actually runs.
Tools & MCP
What happens between an agent asking for a tool and the tool changing something.
Governing AI agent tool calls and MCP tools: access isn't authority
The agent found the right-sounding tool and granted production data access. Why tool visibility isn't permission, why per-call allow and block is only the start, and how to govern agent and MCP tool calls end to end.
See it on one of your own agents
One agent, one consequential action, one target system. We'll walk through identity, approval, durable execution and the record on your own case.