The questions security and engineering teams ask us most. Reach out if yours isn't here.
Security & data handling
No. We never store your source code. Whitebox analysis runs entirely in ephemeral memory: we read it, scan it, and discard it. Nothing persists to disk, so there is nothing to leak, subpoena, or breach.
Credentials are encrypted at the column level under a separate key the database never holds. They're decrypted only inside the sandbox, at the moment of use: never at rest, never written to logs.
Every run is its own isolated environment. Each scan gets a fresh sandbox with no shared file system, memory, or cross-customer state, and it's destroyed on exit.
No. Network egress is locked to your declared target scope at the infrastructure layer. That is a packet-level guarantee, not a policy promise. The agents physically cannot route to anything outside the scope you authorize.
Not yet attested. Our SOC 2 Type I audit is underway with an independent auditor, and Type II follows. In the meantime, the architecture does most of the work the audit checks for: ephemeral sandboxes, locked egress, no stored source. We're happy to share our current security documentation under NDA.
How Xora works
Xora is the layer between found and fixed. Before each deploy, autonomous agents attack your staging environment the way a real adversary would, prove what's genuinely exploitable with a working reproduction, and hand your team the evidence and the fix.
Scanners flag potential issues: thousands of maybes, most of them noise. Xora proves real exploitability by exploiting the flaw in a controlled environment, so every finding comes with a working reproduction instead of a probability score. In practice that means close to zero false positives.
Not to get started. We can run blackbox or greybox against a target you designate. Whitebox (source) access deepens coverage, and when you grant it, that analysis runs in ephemeral memory and is discarded (see above). We never ask for standing repo access up front.
Yes. Add a test account for each user role (an admin and a regular user, say), with its own login URL if it signs in somewhere different. Accounts behind an authenticator app work too: paste the setup key or upload the QR code, which is decoded in your browser. A test-login button confirms each account signs in before a scan. Agents then try one role's access against another's, and never see the passwords: an isolated sign-in step hands them only the session.
Xora runs against the staging or pre-production environment you designate, before code ships, so production is never the test target. Egress is locked to your authorized scope, so testing stays contained.
Yes. Xora is built to run on every deploy with a single flag, returning a clean pass/fail signal and halting the deploy when it finds a working exploit. No new workflow for your engineers, and on-demand assessments typically return same-day results.
Yes. The xora mcp command runs a read-only MCP server that connects Claude Code, Cursor, or any MCP-capable editor to your findings. It scopes to the repository you have open, and for any finding returns the vulnerable code, exploit evidence, attack chains, remediation guidance, and fix-PR link.
Compliance & reporting
For most teams, yes. Xora's reports are auditor-accepted and can stand in for a point-in-time pentest for SOC 2, ISO 27001, and HIPAA. A couple of regimes still require an accredited human assessor (a 3PAO for FedRAMP, a QSA for PCI Level 1); there, Xora provides continuous validation that strengthens your evidence rather than fully replacing the assessor.
Each assessment produces the penetration-test evidence SOC 2, ISO 27001, HIPAA, and PCI DSS ask for: the request, response, and full reproduction steps for every finding, classified by CWE, OWASP Top 10 category, and CVSS severity. Auditors and insurers can read it as-is.
A working proof-of-concept: the exact request and response, reproduction steps, severity, its CWE and OWASP classification, and actionable remediation guidance. Not a CVE guess. A demonstrated exploit your engineers can reproduce and close.
Pricing & engagement
Exactly what it says: on our on-demand plans, if a validated high- or critical-severity exploit isn't found, you don't pay for the assessment. We're confident enough in the depth of our testing to put our fee behind it.
Every assessment is scoped to the application it covers, so pricing scales with your app's complexity rather than a fixed tier. Continuous, every-deploy testing is custom-scoped too. Request a quote and we'll come back with a number, usually the same day.
Both. Run a one-time on-demand assessment whenever you need it, or move to continuous validation that tests every release automatically. New code ships, new tests run.
Still have questions?
We'll walk you through exactly how Xora would test your application — and answer anything security or engineering wants to dig into.