SentinelMCP – An open-source firewall for AI agents that use MCP (github.com)

Technosive Ltd. has announced SentinelMCP, an open-source security gateway designed to protect AI agents using the Model Context Protocol (MCP). Currently in alpha stage, this firewall provides critical runtime security features, including inspection of tool calls, policy enforcement, PII/secret redaction, and audit logging. What sets SentinelMCP apart is its dual deployment architecture; it can operate as a standalone sidecar proxy or as an inline SDK directly within an application. This allows developers to integrate significant security measures with minimal latency—sub-millisecond for inline calls—making it particularly appealing for environments requiring real-time performance. The significance of SentinelMCP lies in its approach to enhancing security for AI agents by enabling human-in-the-loop (HITL) workflows and compliance through robust policy management. Users can define custom YAML-based policies with hot-reloading capabilities, allowing for adaptive risk management. Additionally, it features a built-in Data Loss Prevention (DLP) mechanism that scans for sensitive information, with support for both local and customized regex patterns. By pushing the boundaries of security for AI applications and fostering community input during its development phase, SentinelMCP aims to address pressing vulnerabilities in AI/ML deployment scenarios.