API security

Test the API the way an attacker reads the spec.

Point us at a REST, GraphQL, or gRPC API and we ingest its OpenAPI spec, map every operation, then fuzz each one and hunt shadow routes. The deep business-logic flaws — BOLA/BFLA — get a differential authz test that proves the bypass, never guesses it.

Wraps best-in-class OSS · grounded, low false positives · fixes are human-approved

What we assess

Coverage that maps to real risk.

Spec-driven fuzzing

OpenAPI ingest → schemathesis + nuclei across every operation — injection, auth, conformance, mass-assignment.

Shadow-route discovery

kiterunner finds the undocumented endpoints that aren't in the spec but are live.

BOLA / BFLA, proven

A differential authz test replays the victim's request as the attacker — a hit is verified (2xx with victim data), never a guess.

GraphQL aware

Introspection + per-resolver checks for the queries that leak data or escalate.

Powered by nuclei, kiterunner, schemathesis — best-in-class OSS, wrapped (never re-built in-house), so coverage equals the standalone tool.

How it works

From target to fix, grounded at every step.

1
Ingest the spec

OpenAPI/Swagger → the exact operation inventory becomes the scan surface (no blind crawling).

2
Fan out per method

Each operation gets the right tool — spec fuzz, injection, shadow-route brute — run in the sandbox.

3
Prove authz logic

The consent-gated apiauthz prober demonstrates BOLA/BFLA with two real identities, so a finding is exploitation-proven.

Compliance mapping. API findings map to OWASP API Top 10, SOC 2 (CC6.1), and PCI-DSS (6.2.4) where a control nexus exists.
Three ways to run it

The product, or the product + an expert.

The hard calls — the judgment, the legal attestation, the named accountability — are a human's. The only question is whose.

Frequently asked

Do you need the OpenAPI spec?

It's ideal — the spec gives the exact operation inventory for precise, low-noise fuzzing. Without one we fall back to crawling the base URL.

Can you find BOLA/BFLA (broken object/function-level authorization)?

Yes — that's the differential authz test: it replays one identity's request as another and flags a bypass only on proven access to the victim's data, so the finding is verified.

Is active testing safe?

The active prober is consent-gated and benign-by-construction (canary probes, true/false differentials that extract no data). You authorize it per engagement.

Scan an API in minutes.

Start free, or have our expert run the whole engagement for you. Either way, you get a grounded, audit-ready result — not a noisy report you have to triage.