PENTEST

Continuous pentesting that proves every finding

Point it at a domain, an API, or an internal network. It works the way a pentester works: enumerate the surface, try attacks, chain the ones that land. A scanner matches signatures and hands you a queue to triage. The fox reports only what it could make happen — and it makes it happen twice, once to prove it and once after you shipped the fix.

Read the docs Meet the other three

The assessment view.

See everything one assessment produced: the summary the agent wrote, the findings it proved, and the activity behind them. Open the report from here, or download it as a PDF.

JWT tenant isolation and admin surface validation

Completed4 days ago31 minutes

4Risks1 critical2 high1 medium

ExecutionAutomated

Started4 days ago

Duration31 minutes

Risks4

Credits180

Summary

The scope granted a standard workspace token for a non-privileged member of tenant 4821. The goal was to establish whether that token can reach data or operations belonging to another tenant.

It can. The token was accepted on a tenant-scoped query for account 4822 with no error and no audit entry, and the admin GraphQL catalogue enumerated persisted operations that a standard session should not see. Signup was then shown to trust a client-supplied tenant_id, which is how a caller reaches an arbitrary tenant in the first place.

What was refused

SQL injection against the catalogue filter was refused — the query is parameterised. A JWT alg:none downgrade was rejected by the verifier. Both are recorded because a run that lists only what worked is describing a demo rather than a target.

Recommended order

Derive tenancy server-side at signup and reject client-supplied tenant ids.
Scope the token check to the tenant on every read path, not just on writes.
Record cross-tenant reads in the audit log so the next attempt is visible.