Performance
Benchmarks
Real data. Not marketing numbers. Every benchmark measured on production-grade infrastructure with full governance pipeline active. P50, P95, and P99 — so you know the worst case, not just the best.
Latency Benchmarks
Measured with full governance pipeline active. No caching shortcuts. No synthetic optimizations.
Policy Evaluation
<8msTime to evaluate a single action against all applicable policies. Includes policy lookup, rule evaluation, and decision rendering.
Kill Switch Check
<3msTime to check organization and agent kill switch state. Performed before policy evaluation on every action submission.
Audit Chain Append
<12msTime to compute hash, chain to previous entry, encrypt payload, and persist to storage.
Request Authentication
<2msTime to verify cryptographic request signature including timestamp and nonce validation.
PII Detection Scan
<15msTime to scan action payload for sensitive data patterns including PII, credentials, and API keys.
SDK Overhead
<5msAdditional latency added by SDK per action submission. Excludes network round-trip.
Methodology
All benchmarks run on production-equivalent infrastructure with full governance pipeline active on every measurement. No caching shortcuts. No synthetic optimizations.
Latency measurements exclude network round-trip (measured at the application layer). SDK overhead is measured locally to isolate from network latency. Real-world latency includes network round-trip to the OnLeash API.
Governance at the Speed of Your Agents
Sub-8ms overhead. Zero perceptible latency for your AI agents.
Start Benchmarking