Signal Sentinel - Security for MCP Servers and Agent Skills
25 security rules. Triple OWASP mapping. Scan MCP server configurations and Agent Skill packages before attackers do. Free and open source.
Two growing attack surfaces. One scanner.
The Model Context Protocol has become the universal standard for connecting AI agents to external tools. Meanwhile, Agent Skills (SKILL.md format) are rapidly emerging as the standard for reusable AI instructions - adopted by Claude Code, Codex CLI, Cursor, Windsurf, and 20+ platforms.
Both are growing attack surfaces. Research identified over 2,000 MCP servers operating with no authentication. Community skill marketplaces host thousands of skills, and documented large-scale supply chain attacks have already been observed. Most organisations have no visibility into the security posture of either.
Signal Sentinel addresses both gaps with 25 security rules and triple OWASP compliance mapping.
Three products. One security posture.
Signal Sentinel is a family of security tools released in phases. Each product addresses a different layer of MCP security - from assessment to enforcement to classification.
Sentinel Scanner
Security audit for MCP servers and Agent Skills
CLI tool with 25 security rules scanning MCP server configurations and Agent Skill packages against OWASP Agentic AI Top 10, Agentic Skills Top 10, and MCP Top 10. A-F grading, SARIF output, offline mode, suppression workflow, rug-pull detection. Auto-discovers across 7 platforms. .NET tool, Docker, and source. v2.3.0. Free. Open source. Apache 2.0.
Learn moreSentinel Gateway
Real-time MCP proxy and firewall
Sits between your AI agents and MCP servers, enforcing security policies in real-time. Request validation, tool-level access control, PII redaction, behavioural anomaly detection, and a kill switch for rogue agent behaviour. Full audit logging for compliance. SaaS subscription with free tier. Enterprise and Defence tiers available.
Register interestSentinel Classify
Document classification MCP server
Enterprise-grade MCP server providing automated document classification and sensitivity labelling for AI agents. Deploys across Azure, AWS, Google Cloud, and Oracle. Purpose-built for organisations with regulatory obligations around data handling and classification. Multi-cloud marketplace distribution. UK data residency from day one.
Register interestBuilt by practitioners, not theorists
Signal Sentinel is built by Signal Coding - a UK-based AI-accelerated software engineering company serving Government, Defence, and National Security. Every product is developed under our governed AI development framework, following MOD JSP 440/656 security principles, OWASP guidelines, and Secure by Design methodology.
We built Signal Sentinel because we saw the same security gap in the agentic AI ecosystem that we have spent years addressing in government software delivery - ungoverned tools, no audit trail, no succession plan. The tooling should be as secure as the systems it protects.
Product roadmap
Frequently asked questions
What is Signal Sentinel?
Signal Sentinel is a family of open-source security tools for the Model Context Protocol (MCP) and Agent Skill ecosystems. The current release is Signal Sentinel Scanner v2.3.0 – a CLI tool with 25 security rules mapped to three OWASP frameworks. Signal Sentinel Gateway (real-time MCP firewall) and Signal Sentinel Classify (document classification MCP server) are in development.
How many security rules does Signal Sentinel have?
Signal Sentinel Scanner v2.3.0 has 25 detection rules plus 1 informational rule. 16 rules cover MCP server scanning and 9 rules cover Agent Skill scanning.
Is Signal Sentinel free?
Yes. Signal Sentinel Scanner is free and open-source under the Apache 2.0 licence. No telemetry, no data collection, no registration required. Signal Sentinel Gateway will have a free tier when it launches in Q2 2026.
What OWASP frameworks does Signal Sentinel cover?
Signal Sentinel Scanner maps findings to three OWASP frameworks: OWASP Agentic AI Security Top 10 (ASI01-ASI10), OWASP Agentic Skills Top 10 (AST01-AST10), and OWASP MCP Top 10 (MCP01-MCP10).
Who built Signal Sentinel?
Signal Sentinel is built by Signal Coding Ltd, a UK-owned AI-accelerated software engineering company serving Government, Defence, and National Security. It is developed under a governed AI development framework following MOD JSP 440/656 security principles, OWASP guidelines, and Secure by Design methodology.
What platforms does Signal Sentinel auto-discover?
Signal Sentinel Scanner auto-discovers MCP configurations on Claude Desktop, Cursor, VS Code, Windsurf, and Zed. It auto-discovers Agent Skills on Claude Code, Codex CLI, Cursor, Windsurf, and project-level skill directories.
Ready to secure your AI agent integrations?
Start with Sentinel Scanner - free, open source, and ready to run in seconds.
Last updated: May 2026