MCP Pentesting
As AI agents become more autonomous, securing the Model Context Protocol is critical.
Who this is for
Engineering and security teams preparing for a release, a customer security review, or a procurement questionnaire — and anyone who needs to know what an attacker could actually reach.
Methodology
What We Test
- MCP Server and Client implementations
- Tool definitions, schemas, and metadata integrity
- Context injection and manipulation vectors
- Agent-tool authorization and permission scoping
- Data leakage via context windows
- Protocol-level authentication and encryption
How We Test
We analyze the MCP architecture to identify trust boundaries. We perform "tool poisoning" attacks to see if malicious tools can mislead agents. We fuzz protocol messages to find parsing errors. We test if agents can be coerced into taking unauthorized actions via manipulated context.
What You Receive
- Vulnerability report specific to MCP architecture
- Proof of Context Injection or Tool Poisoning
- Recommendations for secure agent design
- Hardening guidelines for tool schemas
How an engagement runs
-
Scope & authorisation
We agree exactly what is in scope, what is explicitly out, and the conditions under which we stop. Nothing starts without written authorisation from someone able to give it. NDA first if you want one.
-
Access & prerequisites
We tell you up front what we need — test accounts at each privilege level, a non-production environment where that applies, documentation, and a named technical contact. Missing prerequisites are the usual reason assessments slip.
-
Reconnaissance & threat modelling
We map the real attack surface and decide which paths are worth the time, based on how your system is actually built rather than a generic checklist.
-
Manual testing & validated exploitation
Testing is hands-on. Findings are reproduced before they are written up, so you are not sent unverified scanner output. Critical issues are raised as soon as we confirm them rather than held for the report.
-
Report & walkthrough
You get an executive summary your board can read and technical detail your engineers can act on — including reproduction steps, what we tested that held, and what was out of scope. We walk your team through it.
-
Retest
Once you have fixed the findings we retest them and confirm in writing what is closed. Retest scope and window are agreed in your scope document before the engagement begins.
Toolkit
- Custom MCP Fuzzers
- Proxy Tools
- Agent Simulators