Agent Safety & Trajectory Benchmark
Agents Tested127
Tool Calls Analyzed48,291
Dangerous Actions Blocked3,817 / 3,842 (99.3%)
Verify Every Agent Outcome with Cryptographic Proof — live benchmarks, attack resilience, and immutable evidence chains for investor due diligence.
Open Live Trust HubInvestor-grade, verifiable runtime metrics — agent safety, accuracy, sub-10ms intercept latency, and cryptographically signed evidence chains.
*Tested across 500+ MCP attack scenarios & multi-agent execution graphs. Open-source benchmark harness available on GitHub.
Prove It — Live Attack → Block → Evidence
Agent Attack Attempt
Unauthorized shell command: rm -rf / — PII exfiltration via read_customer_records