Sentinel Model Context Protocol
Three-stage safety guardrail agent for LLM coding assistants (Claude Desktop, Cursor, CodeX), shipped as a plug-and-play MCP plugin.
What it does
Three-stage safety guardrail agent for LLM coding assistants (Claude Desktop, Cursor, CodeX), shipped as a plug-and-play MCP plugin.
Why it needed building
LLM coding assistants can propose file edits, shell commands, and network calls directly — and most setups have nothing standing between the model's suggestion and its execution.
How it was engineered
Designed Sentinel as a three-stage review gateway that sits in the Model Context Protocol path between an agent and its execution environment, scoring and gating every proposed action before Claude Code, Cursor, or CodeX are allowed to run it.
rm -rf / --no-preserve-rootBLOCKEDWhere it landed
Across 100k test payloads Sentinel logged zero unauthorized escapes, while holding gateway latency under 5ms across 40+ concurrent agent registries — safe without being slow enough to get disabled.