Six attack surfaces. Deterministic, reproducible, no LLM in the hot path. Each finding comes with a rule ID, severity, the exact payload we used as evidence, and a remediation hint.
Six attack surfaces. 27 deterministic checks per scan. Each finding comes with a rule ID, severity, the exact payload we used as evidence, and a remediation hint — no LLM in the hot path, no black box.
Real payloads against your live server — but we never write, delete, or persist anything. Path-traversal probes read; we don't mutate state.
We speak JSON-RPC 2.0 natively, enumerate every tool, classify by capability, then fire targeted payloads at each parameter that looks risky.
If a probe can't run (timeout, blocked, no response) we tell you. Operational findings are separate from security findings and never affect your score.
The formula is weighted by severity and published. No black box, no “security score” marketing points — just a transparent rubric.