Services
Agentic AI Security Review
A three-week technical review of your agent, MCP, and RAG deployment: attack surfaces, vulnerabilities, and gaps documented before they become incidents. €18,000 fixed.
Contributor, OWASP Top 10 for Agentic Applications 2026 · CISSP · PhD · 15+ years enterprise security
What I assess
Prompt injection, data leakage, jailbreaks, multi-turn manipulation — the full attack surface of your LLM deployment.
Tool use boundaries, privilege escalation, confused deputy attacks — security for agents that take actions.
Authentication, data isolation, logging, secrets exposure — the foundation your LLM runs on.
Third-party models, RAG data sources, fine-tuning data, vector store security — the risks in your stack.
How it works
Discovery call
I learn about your system, your stack, and what you're most concerned about.
Scoping
Clear scope document before anything starts. Timeline, access requirements, deliverables.
The Review
Three weeks. I work with your team to get the access I need without disrupting operations.
Report & walkthrough
Written findings delivered, followed by a live session with your team.
Pricing
One fixed price for the standard three-week Review.
Agentic AI Security Review
Three weeks, fixed scope
- Your agent, MCP, and RAG deployment
- Three-week turnaround
- Written findings report, OWASP ASI01–ASI10
- Prioritized remediation roadmap
- Live walkthrough session
- 90-day free re-test
Larger estates — more than four connected agents, multi-tenant deployments, or several environments — are scoped and quoted on the call. Prices exclude VAT; invoice issued by Molnsys AB.
Common questions
Yes. Mutual NDA available before kickoff. The discovery call itself is enough to scope without one.
Pre-deployment is when assessment is most valuable. The method is built for systems still in active build, where findings are cheap to act on.
I don't access production data. Assessments run against staging environments or synthetic data your team provides.
That's the point. Remediation guidance and the 90-day re-test are included so the engagement ends with a safer system, not just a list of findings.
Yes. The methodology section is designed to satisfy procurement reviewers asking for evidence of independent third-party AI security assessment.
€5M professional liability coverage. Certificate available on request.
Because the methodology is repeatable and I'm not running an audit firm. The price reflects what the work actually takes, not what the market will tolerate.
Why this matters
Most teams don’t know where their AI system is vulnerable until a third party finds it. By then, the damage is done and the remediation is expensive.
The Agentic AI Security Review gives you a structured, technical review of your deployment before that happens. Three weeks, one fixed price. You get a written report with every finding documented: severity level, how to reproduce it, and specific remediation guidance your engineers can act on directly.
Methodology
Every Review follows the Molntek Agentic AI Security Methodology: a structured framework mapped to the OWASP Top 10 for Agentic Applications 2026 (ASI01 to ASI10), with cross-references to NIST AI RMF, MITRE ATLAS, and EU AI Act Article 15 in the deliverable. The same methodology runs every engagement, which is why the price is fixed.
What’s included
A written findings report with severity rankings, reproduction steps, and a prioritized remediation roadmap. Delivered with a live walkthrough session for your engineering and security teams.
The Review includes a 90-day free re-test if you implement the recommended remediations. The goal is a safer system, not a finding count.
Larger estates — more than four connected agents, multi-tenant deployments, or several environments — are scoped on the call and quoted separately. The €18,000 covers the standard three-week Review.
See an example report
A sanitized sample report from a synthetic engagement is available to download. It shows the full report structure, severity ranking format, OWASP Top 10 for Agentic Applications 2026 mapping, and remediation roadmap layout you would receive at the end of a real engagement.
Want to see the deliverable first?
Download a sanitized sample assessment report. No call required.
Sounds like a fit?
A short call is usually enough to figure out whether this is what you need and what it would look like.