About MCP Sentinel / Evidence before trust
A clearer picture
of the tools you trust.
MCP Sentinel connects server identities, security findings, runtime observations, and assessment history. It helps you inspect the evidence behind a deployment decision.
What ran, what was observed, and what remains unresolved belong in the same picture.
The product's purpose
Tool access deserves a closer look.
MCP servers connect agents to files, databases, APIs, and code execution. A useful security assessment needs to explain both the risks it found and the limits of what it could examine.
The registry makes findings discoverable. Research surfaces examine how capabilities combine. Reports connect those conclusions to evidence, coverage, and assessment history, so a decision can be inspected rather than reduced to a badge.
One product. Several ways to see.
The registry
Find server identities and inspect the findings, assessment method, coverage, and history already on file.
Listed does not mean completely assessed.
Browse server reportsCross-server research
Explore capability combinations and multi-step patterns with the nature of their supporting evidence clearly labelled.
Observed, inferred, and hypothetical remain distinct.
Explore the connectionsReports & verification
Place findings in a framework context and review the available signed assessment reports alongside their scope and limitations.
A signature protects a record; it is not a safety guarantee.
Explore assessment reportsFrom identity to assessment
How an assessment is assembled
Four layers of the assessment pipeline, summarized here. Source acquisition, sandbox execution, and cross-server analysis each have their own boundaries. The deterministic rule engine judges the available evidence; the report retains what was not observed.
Discovers server identities from package registries, repository sources, and MCP registries. Records the discovery source and errors; a listing alone is not a completed assessment.
Acquires an artifact or connects through the MCP SDK for read-only enumeration. Live third-party endpoints are never used for tool invocation. A fetched artifact may be exercised only inside a sandbox that passes the execution boundary checks.
Evaluates the 184 detection rules against the available source, declarations, dependencies, protocol metadata, and eligible runtime observations. Each finding carries evidence and remediation. Coverage records the inputs and execution gaps.
deriveVerdict() turns the findings and the coverage ledger into one of four verdicts. Risk: any critical finding, or a lethal trifecta. Caution: any high finding. Safe: no critical or high finding AND coverage sufficient for that verdict. Otherwise Insufficient coverage - never Safe by default.
184 detection rules across 21 categories
Every rule is a typed TypeScript implementation with a mandatory evidence chain. The engine is deterministic. Every finding requires rule_id, evidence, and remediation. Categories A-K are detailed below; the full A-U catalogue is on the taxonomy page.
Description Analysis
9 rulesRequires: Tool metadata
Unicode homoglyph attacks, zero-width character injection, base64-encoded payloads hidden from human review
Schema Analysis
7 rulesRequires: Tool metadata
Parameter-level injection surface, dangerous defaults (recursive:true, disable_ssl_verify:true), unconstrained additional properties
Code Analysis
17 rulesRequires: Source code
20+ secret token formats, prototype pollution, unsafe deserialization, JWT algorithm confusion, timing attacks on secret comparison
Dependency Analysis
7 rulesRequires: package.json / requirements.txt
50+ confirmed malicious package names including MCP ecosystem typosquats, Levenshtein distance for typosquat detection, dependency confusion (high version number trick)
Behavioral Analysis
7 rulesRequires: Connection metadata
Live connection checks - transport security, authentication presence, response time anomalies, tool count explosion - plus what the sandbox observatory witnessed a tool actually do versus what it declared
Ecosystem Context
8 rulesRequires: Tool metadata
Lethal Trifecta (private data + untrusted input + external comms = a Risk verdict), circular data loop enabling persistent injection, multi-step exfiltration chain
Adversarial AI
7 rulesRequires: Tool metadata + scan history
Attacks that only work because the target is an AI - indirect injection gateways, trust assertion injection, tool behavior drift (rug pull), DNS data exfiltration
2026 Attack Surface
3 rulesRequires: Mixed
March 2026 additions: OAuth 2.0 MCP auth flaws, injection into MCP initialize response fields (processed before tool descriptions), cross-agent propagation in multi-agent orchestration
Protocol Surface
17 rulesRequires: Protocol metadata + annotations
MCP protocol primitive attacks: annotation deception (readOnlyHint lies), sampling abuse (23-41% attack amplification), cross-config lethal trifecta, consent fatigue exploitation (84.2% success rate), credential harvesting via elicitation
2026 Threat Intelligence
9 rulesRequires: Source code + tool metadata
CVE-backed rules from real-world incidents: cross-agent config poisoning (CVE-2025-53773), git argument injection (CVE-2025-68143), full schema poisoning (CyberArk FSP), health endpoint disclosure (CVE-2026-29787), tool output poisoning (CyberArk ATPA), OpenAPI spec injection (CVE-2026-22785)
Compliance & Governance
20 rulesRequires: Source code
Framework-mapped rules closing gaps across ISO 27001, ISO 42001, EU AI Act, NIST AI RMF, CoSAI, MAESTRO - audit trail integrity, human oversight, credential lifecycle, supply chain hooks, sandbox enforcement
OWASP MCP Top 10 mapping
These selected rule mappings provide an orientation to the OWASP MCP Top 10. A server's posture shows whether findings were detected in each category; an empty category is meaningful only alongside its coverage. The full active catalogue is available on the taxonomy page.
| ID | Category | Rules |
|---|---|---|
| MCP01 | Prompt Injection | A1, A5, A7, A8, A9, B5, F1, F6, G1, G2, G3, H2, I3, I6, I7, J3, J5, J6 |
| MCP02 | Tool Poisoning | A2, A4, A6, F2, F5, G5, I1, I2, J5, J6, K12, K13 |
| MCP03 | Command Injection | C1, C9, C13, C16, J2, J7 |
| MCP04 | Data Exfiltration | A3, F1, F3, F7, G7, I9, I13, K18 |
| MCP05 | Privilege Escalation | C2, C8, C10, C12, H1, I4, I12, J1, K6, K8, K14 |
| MCP06 | Excessive Permissions | A2, B3, B7, E4, F2, I11, I16, K6 |
| MCP07 | Insecure Configuration | B6, C7, C8, C11, C14, C15, D6, E1, E2, H1, I15, J4, K7, K17, K19 |
| MCP08 | Dependency Vulnerabilities | D1, D2, D3, D4, D5, D6, D7, K9, K10 |
| MCP09 | Logging & Monitoring | C6, E3, K1, K2, K3, K20 |
| MCP10 | Supply Chain | D3, D5, D7, A4, F5, G6, I5, J7, K9, K10, K11 |
The verdict
How it is derived
Risk any critical finding,
or a lethal trifecta (F1 / I13)
Caution otherwise, any high finding
Safe no critical or high finding
AND coverage sufficient for that verdict
Insufficient no critical or high finding,
coverage but too little was checked
to call it safe
Never Safe by default.What it means
If a server reads private data AND ingests untrusted content AND has external network access — or a client config spreads those three across servers — the verdict is Risk regardless of other findings. No amount of clean code compensates for this combination.
Findings are positive evidence: a supported critical finding is Risk at any coverage. The absence of findings is not. A clean result is Safe only when the coverage band is high or medium and no high or critical rule was left with an unanswered input — otherwise it is Insufficient coverage. Refused, out-of-scope and not-observed are never “clean”.
Per-finding confidence, per-framework compliance percentages and the attack-graph exploitability score (0–1) are separate signals and remain. There is no composite number to fold them into.
Architecture principles
All 184 detection rules are TypeScript TypedRuleV2 implementations registered in a central registry - metadata in the registry, detection in code, no YAML. Adding a rule never requires changing engine code.
The detection engine uses typed rules, AST analysis, capability graphs, entropy, structural parsing, and vulnerability lookup. Scoped assisted framework and observation workflows do not replace its evidence requirements.
Crawlers store all raw metadata. Analysis is a separate pass. We never discard data because we don't have a rule for it yet.
Every scan result is immutable. Verdicts and findings change over time; the history table tracks every change. Trends are first-class.
Live third-party endpoints receive read-only enumeration. Runtime checks may exercise our fetched copy of an artifact only after the sandbox verifies denied network egress and filesystem confinement. Refused or failed execution remains a gap.
Every finding must include rule_id, evidence (what triggered it), and remediation (how to fix it). Findings without evidence are useless.
02 / Preserve the record
An assessment has a scope.
Its record should carry it.
The registry keeps recorded findings and assessment history. Reports carry those findings with their method and coverage, so the evidence remains attached to the conclusion.
Signed report availability depends on the deployment's signing configuration. Failed checks and missing observations remain questions to resolve.
A signature protects the report under its signing scheme; it does not guarantee a server's safety.
Explore report integrityThe principle we return to
Make the evidence inspectable.
A failed check is not a clean result. An inferred path is not a witnessed attack. A report is strongest when its findings and its unanswered questions can both be examined.