impeccable-mcp
MCP server that wraps impeccable's design-lint engine to scan URLs and raw HTML for 59 deterministic anti-patterns, returning structured findings grouped by slop vs quality via headless Chrome.
README
<img src="assets/fathom-header-banner.svg" alt="Fathom Works — impeccable-mcp" width="100%">
$ impeccable-mcp
An MCP server wrapping impeccable's design-lint engine — headless Chrome plus 59 deterministic anti-pattern rules — so any agent can scan a URL and get back structured findings, no per-agent setup required. No LLM, no API key needed for the scan itself; the engine is pure detection.
[ what it does ]
- Takes
impeccable's public library exports (detectUrl,detectHtml,ANTIPATTERNS,createBrowserDetector) — an official npm dependency, not a fork — and exposes them as MCP tools - Runs one long-lived headless Chromium instance, each scan in its own incognito context
- Serves MCP over streamable HTTP on port 8000, same shape as gsd-browser-mcp
- Lets any MCP client (Claude Code, Fred, Billy, whatever's next) scan a website's design without local setup
[ tools ]
| Tool | Needs a browser | What it does |
|---|---|---|
scan_url |
yes | Loads a URL in headless Chrome, runs all detector rules, returns findings grouped by slop (AI design tells) vs quality (a11y/design issues). Accepts optional headers / cookies for authenticated pages. |
scan_html |
no | Runs the static detector against a raw HTML/CSS string. No browser launched — use this to check generated markup before it's written to disk. |
screenshot |
yes | Returns a PNG of the rendered page as MCP image content. |
list_rules |
no | Returns the full rule registry (id, name, category, description) from impeccable's ANTIPATTERNS. |
Header/cookie values you pass to
scan_urlwill appear in the calling agent's transcript. That's a deliberate tradeoff for reaching authenticated pages, not an oversight — seeCONTRIBUTING.md. Don't pass anything you wouldn't want logged.
The private-target guard
scan_url and screenshot refuse RFC1918, loopback, and link-local targets by default. This
service's whole job is "fetch whatever URL an agent hands it" — without the guard, a
prompt-injected agent could turn it into a LAN port scanner or point it at Portainer/Home
Assistant from inside the trusted network. Set IMPECCABLE_ALLOW_PRIVATE=1 only if you specifically
need to scan an internal site and understand the tradeoff.
[ quick start ]
1. pull and run
$ docker run -d \
-p 8000:8000 \
--shm-size=512mb \
-e IMPECCABLE_MCP_TOKEN=$(openssl rand -hex 32) \
ghcr.io/jemplayer82/impeccable-mcp:latest
2. connect your mcp client
http://localhost:8000/mcp
with Authorization: Bearer <IMPECCABLE_MCP_TOKEN>.
[ docker compose ]
services:
impeccable-mcp:
image: ghcr.io/jemplayer82/impeccable-mcp:latest
restart: unless-stopped
shm_size: "512mb"
mem_limit: 1536m
init: true
ports:
- "8000:8000"
environment:
IMPECCABLE_MCP_TOKEN: your-token-here # pragma: allowlist secret — doc placeholder
healthcheck:
test: ["CMD", "curl", "-sf", "http://localhost:8000/healthz"]
interval: 30s
timeout: 5s
retries: 3
start_period: 20s
[ environment variables ]
| Variable | Required | Description |
|---|---|---|
IMPECCABLE_MCP_TOKEN |
Yes | Bearer token clients must send. Server refuses to start if unset. |
PORT |
No | Listen port inside the container. Default 8000. |
IMPECCABLE_ALLOW_PRIVATE |
No | Set to 1 to allow scans of RFC1918/loopback/link-local targets. Default off. |
MAX_CONCURRENCY |
No | Max concurrent browser-backed scans. Default 2. |
BROWSER_IDLE_TIMEOUT_MS |
No | Close the idle browser after this long with no scans. Default 300000 (5 min). |
[ how it works ]
A single Node/ESM process:
@modelcontextprotocol/sdk'sStreamableHTTPServerTransportover rawnode:http— stateless, a fresh server + transport per request, no session bookkeeping.src/scan.mjsholds one Chromium viaimpeccable'screateBrowserDetector, relaunching on crash and closing afterBROWSER_IDLE_TIMEOUT_MSof inactivity. Each scan gets its own incognito browser context so headers/cookies from one call never leak into another.- Findings come back in
impeccable's stable shape (antipattern,name,description,severity,category,snippet), grouped by category with a count summary.
impeccable itself needs no LLM and no API key — it's a deterministic rule engine. This server
doesn't add one either; it's transport, auth, browser lifecycle, and a URL guard around a library.
[ related ]
- impeccable — the underlying detection engine (don't
file issues/PRs there from this project; see
CONTRIBUTING.md) - gsd-browser-mcp — the sibling MCP this repo's transport, auth, and Dockerfile pattern were cloned from
- Impeccable Chrome extension — same rule engine, for a human looking at DevTools directly. This server exists for agents; the extension is still the right tool for eyeballing a page yourself.
[ license ]
Apache License 2.0 — see LICENSE.
<img src="assets/fathom-footer-banner.svg" alt="Fathom Works — sound the depths before you set a course" width="100%">
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.