Use cases
Task-focused walkthroughs: real policies, real agents, real resources — from read-only Jira access for Claude to per-team token spend.
SecureAuth Agent Authority's concepts and resource guides explain how the pieces work; these walkthroughs start from a real question a security or platform team asks and end with a policy rule, a scope, or an Analytics view you can copy into your own gateway. They fall into two jobs: controlling what agents can do, and seeing what they did.
Control access
Give Claude read-only access to Jira
Allow every read tool, block every write, with one deny-by-default rule. Pattern: allow-list.
Block an agent from deleting GitHub branches
Let an agent run the full pull-request workflow while denying the tools that merge or overwrite branches. Pattern: deny destructive actions.
Sandbox an agent to one Slack channel
Scope Slack access to a single channel ID by checking the channel each call targets. Pattern: argument-level condition.
See activity
Audit everything an AI agent did
Go from "which agent touched this" to a specific actor, tool call, and policy decision in Agent Trail. Pattern: investigation.
Track AI token spend by team
Break down Claude and Codex token cost by team with Group filters, then pull the matching tool-call activity through the same lens. Pattern: cost attribution.