Inspector MCP server

Expose protect/report as MCP tools for agents.

Warning

Creytix Inspector MCP server is in active development — here's what it will do and what's live so far.

Overview

Inspector's checks are useful to a person reading a report — and they should be just as usable to an AI coding agent working on your site. The Inspector MCP server is meant to expose Protect's rule checks and Inspector's reports as tools an agent can call directly, so an agent fixing a page can check its own work against the same rules a person would use before shipping.

How it works

The target flow: an MCP-compatible agent connects to the Inspector MCP server, calls a check — for example, "run Protect against this page" — as a tool, and gets a structured pass/fail result back: the same result a person would see in Inspector, just callable in-line during agent work.

Steps

•
Connect an MCP-compatible agent
Point your agent's MCP configuration at the Inspector MCP server.
•
Call a check as a tool
The agent invokes a Protect or report check the same way it would call any other tool.
•
Get a structured result back
Pass/fail results return in a form the agent can act on directly — for example, fixing a flagged issue and re-checking.
•
Loop until clean
The agent can re-run checks after each fix, closing the loop before handing the change back to a person.

Capabilities

MCP tool exposure — Protect and report checks callable as tools by any MCP-compatible agent.
Structured results — pass/fail responses shaped for an agent to act on, not just for a person to read.
Target: closed-loop agent QA — an agent checking and re-checking its own work before it reaches a person.

Limits & honest scope

The Inspector MCP server is actively being built. Today: