What is BobSentry?

AI agents can propose consequential actions against real systems. Permissions and orchestration alone do not decide whether a specific action should execute now. BobSentry provides a runtime authorization decision—ALLOW, REQUIRE_APPROVAL, or BLOCK—with signed evidence before execution.

Your agent or gateway intercepts a proposed action, sends normalized action metadata to the Runtime Evaluate API, and receives a decision before calling the external system. Approvals, when required, are bound to that exact proposed action.

ALLOW (proceed), REQUIRE_APPROVAL (hold for operator authorization), or BLOCK (stop before execution).

Normalized action metadata. Sensitive payloads, source code, PHI, credentials, and customer data can remain in the customer environment.

Permissions describe what an agent can access. BobSentry authorizes whether this action should execute now, given policy, risk, target, and context.

Orchestration frameworks schedule and chain agent steps. BobSentry is an independent authorization boundary immediately before an action reaches a real system—not a workflow engine.

Observability records what happened. BobSentry decides whether an action may happen, and produces signed evidence of that decision.

When the decision is REQUIRE_APPROVAL, an operator authorizes or rejects the exact proposed action. Approvals should bind to action metadata, target, policy version, and expiration.

Any consequential agent action you can intercept before execution—external messaging, financial operations, EHR writes, file and Git changes, dependency installs, shell commands, infrastructure applies, and deployments.

Founders and engineers building action-taking AI agents, infrastructure teams deploying coding agents, and security architects responsible for agent execution in regulated or high-impact environments.

Create a sandbox account at the BobSentry application, generate an API key, call the Runtime Evaluate API, and inspect signed evidence. See the Developers page for a five-minute path.