Explore access paths
Access Explorer shows relationships between assets: an agent uses a tool, a server authenticates as an identity, an endpoint receives data, or an asset is defined in a repository.
Focus an investigation
Section titled “Focus an investigation”- Search for the agent, server or identity you want to understand.
- Focus that asset and choose a useful hop limit.
- Keep evidence-state filters explicit. Inspect configured and observed edges separately before combining them.
- Select an edge or node to inspect its source evidence.
- Open the related asset or finding when you need more detail.
The hop limit controls how much neighboring context is shown. A large graph can obscure the question; start close to the asset and expand only as needed.
Read an edge precisely
Section titled “Read an edge precisely”An edge labeled can read is different from sends data to or uses tool. Inspect its evidence state and source. A configured permission may be constrained by a deployed identity policy that ModelFence has not tested. An observed call demonstrates a reported interaction, not a complete catalog of effective permissions.
Use the accessible table representation for keyboard navigation, screen readers or a linear review. It mirrors the graph relationships and is often easier to compare against a source configuration.
Share the investigation
Section titled “Share the investigation”Copy a link to preserve the available focus and filter context, or export a PNG when supported. An image is a snapshot: include the organization, evaluation time and evidence context in your review notes. It does not grant recipients access to underlying evidence.
Current limits
Section titled “Current limits”The explorer does not implement an automatic path-between-two-selected-nodes workflow or environment clustering. Expansion is controlled through the hop limit rather than per-node expansion. The graph cannot prove that absent edges are impossible when sources or runtime telemetry are incomplete.