Technosive Ltd. released SentinelMCP, an open-source security gateway that enforces policies on AI agent tool calls in real time, available now in Alpha as both a proxy and a Go library.

What happened

SentinelMCP is a security enforcement engine for the Model Context Protocol (MCP) that inspects and controls what tools AI agents can invoke. It can run as a standalone proxy (working with any programming language) or as an inline Go library, and includes policy enforcement, PII/secret redaction via regex patterns, human approval workflows, and audit logging.

Why it matters

As AI agents gain the ability to call external tools and modify systems, organizations need guardrails to prevent misuse or data leaks. SentinelMCP sits between agents and their tools to block, redact, or pause risky calls—letting teams enforce consistent security rules without rewriting their agent code.

What to watch

The software is currently in Alpha (v0.1), and the Inline SDK mode (which offers sub-millisecond latency with 19μs p99 overhead on the Allow path) shipped in v0.2.0. The project is open-source and actively seeking early adopters and feedback; it is built in Go with hot-reloadable YAML policies and integrates with OpenTelemetry for SIEM pipelines.