Add-on · Agent Safety Readiness Review

Know how your AI agent scores before the enterprise buyer asks.

Enterprise buyers are starting to ask AI agent vendors to prove safeguards live: the agent's own login, limits enforced in software, a stop switch the buyer controls, and a full exportable record. We run your product through all 20 requests in the public MACH Alliance AI agent safety vendor questionnaire first, on a test account, and help you close the gaps.

From $4,500. Most useful after an Integration Sprint, and available on its own.

What buyers check: 20 requests

The questionnaire (September 2026) says slides, videos and screenshots count as a fail, roadmap-only features count as a fail, and a weak answer on any deal-breaker pauses the evaluation.

Demonstrate live

9

  • A full task run
  • Agent login and access
  • A blocked limit
  • Stopping running work
  • Undo, handoff, injection test, approved limit change, full export
Provide

7

  • AI providers and training commitment
  • Model change notice
  • Test results for your setup
  • 90-day monitoring report
  • Incident notice, retention, AI assessments
Answer in writing

4

  • What the agent remembers
  • AI disclosure and challenges
  • Other agents and tools called
  • Usage cost limits
Deal-breakers

5

  • Shared or human login
  • Limits by instruction only
  • Stop needs the vendor
  • Record cannot be exported
  • AI providers not named

What you get

Planned at 2 to 3 weeks, confirmed in the written scope.

  • All 20 requests scored Good, Weak, Missing or Not applicable
  • A recorded 90-minute mock buyer session covering the 9 live demonstrations
  • A fix plan for every deal-breaker, with owner and effort
  • Flint Approve installed on your riskiest agent action, so the login, limit, stop and export deal-breakers can be shown live
  • Draft answers for the 7 documents and 4 written questions
  • An evidence pack and a readout with your team

What you provide

Light on your team's time.

  • A test account set up like your target buyer's deployment
  • One engineer for the intake and the mock session
  • Existing documents: security page, subprocessor list, DPA, any audit reports
  • Your buyer's planned-use answers, if you have them
Agent Safety Readiness Review
From $4,500
Fixed price in the written scope
Book a 20-minute call

Honest terms

What this is, and what it is not.

No pass guarantee

The buyer decides. We show you exactly where you stand and what a buyer is likely to accept. If we find nothing to fix, the report says so.

Test accounts only

We work on your test account under a written scope. We never touch live systems without your written authorization.

Independent

Flinttech is not affiliated with or endorsed by the MACH Alliance. We are not auditors and issue no certification. Flint Approve is early-stage software without an independent security audit.

Expecting an enterprise security review?

Tell us your product and the kind of buyer you are selling to. We will tell you on the call whether the Review is worth it for you yet.

Framework source: MACH Alliance, Vendor Questionnaire: AI Agent Safety, September 2026 (public PDF).