Dynamic Agent Context
Dynamic agent context is an access model in which an AI agent gets the specific tools and data a task needs, authorized on every tool call rather than granted once and carried forward. The agent's context boundary is set by policy at each call, not by a standing grant.
Why It Matters
Zero standing permissions describes what an agent holds, which is nothing. Dynamic agent context describes how each piece of access gets decided. As the agent works through a task, every tool call is authorized against policy on its own. What the agent can access at any moment is therefore the product of those decisions rather than a grant it was handed at the start.
This is not a theoretical distinction. In practice, an agent working on a code review needs read access to a repository and write access to a PR comment. Five minutes later, the same agent working on a different task needs access to a project management tool and nothing else. Dynamic agent context means the agent's permissions are a function of the current task, not a static role assignment.
The term positions the governance layer as an enablement mechanism, not a restriction. The agent gets more context (and therefore more capability) precisely because there is a system managing what it can access and when.