BitssyBytes logoAgent Safety Score

Directory / Amazon Bedrock AgentCore

B

Amazon Bedrock AgentCore

Enterprise agent platform · AWS · 5/7 verified · Last checked 2026-10-16

Cedar policies check every tool request that goes through AgentCore Gateway. A token vault stores OAuth tokens and API keys encrypted with KMS. Code runs in a sandboxed Code Interpreter.

Overall: B (2.40 avg, 5/7 controls verified)

Lock this down first

Attach a policy engine to every Gateway and keep credentials in the AgentCore Identity token vault.

Scorecard

CriterionRatingNotes
Permission granularity3/3Policy in AgentCore evaluates each Gateway request against Cedar policies before allowing tool access.
Human approval gatesUnverifiedNot confirmed in the docs we checked.
Sandboxing2/3Code Interpreter runs code in sandbox environments.
Secrets handling3/3Token vault encrypts OAuth tokens and API keys at rest and in transit, with customer- or service-managed KMS keys.
Audit logs2/3CloudTrail records Gateway API calls and who made them, and harness invocations get tool-call traces in CloudWatch with no extra configuration. Gateway data events are opt-in; no documented log lock.
Kill switch / interrupt2/3StopRuntimeSession terminates a session and stops streaming responses.
Network egress controlUnverifiedNot confirmed in the docs we checked.

Unverified: Human approval gates, Network egress control. That means we could not confirm it in public docs, not that it is missing.

Sources

Vendor? Request a correction. We review corrections within 7 days.