https://protocolprobe.com/mcp/servers/io.github.jakcmencz/sentinel-ir-mcp#reactions

The 30-second version

I recently noticed something interesting on ProtocolProbe.

Sentinel-IR-MCP now meets all 9 of 9 applicable record checks.

Nice number, isn't it?

But here's the interesting part: this isn't a security certificate, a penetration test, or a declaration that an AI agent is suddenly invincible.

It's something more practical.

It means that important parts of a project's public identity and supply chain can actually be checked. And for developers building AI agent systems, that matters more than another shiny badge.

Why should developers care?

Think about how we build agent systems today.

We connect models to MCP servers, packages, APIs, repositories and external tools. Each connection introduces another piece of software that we are trusting.

And sometimes we connect things faster than we check what is behind them.

It's a little like giving someone keys to your office because they have a nice business card. The card might be genuine, but perhaps we should check who issued it first.

This is where transparent records become valuable.

ProtocolProbe checks nine public signals, including:

  • Is the registry entry active, and does it explain how to run the server?
  • Is the source repository public and maintained?
  • Is there a published security policy?
  • Is the licence clearly identified?
  • Is the listed package version available and current?
  • Does the npm package have signed build provenance?
  • Are there known critical or high-severity security advisories?

These are not glamorous features. Nobody will put a movie trailer together for an SPDX licence identifier.

But they help answer a much more important question:

Can another developer independently verify the basic health and origin of the software they are about to connect to an agent?

For Sentinel-IR-MCP, the current result is 9/9. It also appears at #121 among 1,275 packages meeting all nine checks, ranked by monthly downloads.

That's useful external visibility. Not because a ranking magically makes software better, but because verifiable information makes it easier for developers to compare projects and make informed decisions.

Verdict
Sponsor Sentinel IR development here: https://github.com/sponsors/jakcmencz?o=syg&sc=r
Source (GitLab, Apache-2.0): https://gitlab.com/jakubrojicek11/sentinel-ir
npm: https://www.npmjs.com/package/sentinel-ir
Playground (runs in your browser): https://sentinel-ir-playground-dc30c8.gitlab.io/
MCP Server on npm: npmjs.com
https://glama.ai/mcp/servers/jakcmencz/sentinel-ir-mcp
By "jackymenCZ" .(https://dev.to/jackymencz)

What does this mean for Sentinel-IR-MCP?

For me, the important part isn't the number itself.

It's that the project has publicly verifiable foundations: an accessible source repository, an active release, a security policy, a declared licence and signed build provenance.

These details are part of the infrastructure of trust.

Sentinel-IR itself focuses on giving AI agents structured information about code: risk signals and useful context instead of simply handing over raw source.

ProtocolProbe looks at a different layer: the public records surrounding the MCP server and its package.

Those two perspectives complement each other.

One helps an agent understand the code it is inspecting. The other helps developers understand the software they are considering connecting.

And there is an important limitation worth mentioning: ProtocolProbe did not start Sentinel-IR-MCP or call its tools because a required configuration setting was missing. So the 9/9 result is not a runtime security assessment.

That's a distinction I want to keep clear.

A good checklist is not a force field. But a project that makes its foundations visible gives developers something real to inspect.

The bigger picture

As AI agents become more capable, choosing their tools will become just as important as choosing their models.

Developers should not have to rely entirely on descriptions written by the people publishing those tools.

They should be able to inspect the evidence.

And perhaps that's the real value of a result like 9/9.

Not a promise that nothing can go wrong.

Just nine fewer questions left unanswered.

I'll take that over another AI-generated badge saying "100% secure" any day.

Project: "Sentinel-IR-MCP" (https://protocolprobe.com/mcp/servers/io.github.jakcmencz/sentinel-ir-mcp)

ProtocolProbe record: "9/9 record checks" (https://protocolprobe.com/mcp/servers/io.github.jakcmencz/sentinel-ir-mcp)

Checked November 2026. Record checks reflect public information, not a security guarantee or runtime test.

Copyright © 2026 jackymenCZ. All rights reserved. Written for the community.