pixelgrade

pixelgrade lets you check whether the advertising pixels on a page actually work, without opening a browser yourself. Give it a landing page and it grades the Meta, TikTok and OpenAI pixels it finds there: whether they wait for cookie consent, whether the same event fires twice, whether event IDs line up between the browser and any server copy, and whether the click IDs that ads append — fbclid becoming _fbc, for example — are captured. When something is wrong you get the finding, an explanation, and a step-by-step guide for fixing it on your platform.

The package itself checks nothing locally. It is a thin bridge to the hosted pixel-check service: it answers the tool list from a verbatim capture of that service and forwards every tool call upstream, returning the reply untouched. All four tools are read-only. Checking reads a page you name; reading a report, listing findings and reading a fix guide change nothing, install nothing and touch no ad account.

Tools

• check_page — check a page's pixels and grade each one. Required: url (the page to check, usually the landing page the ads point to, 3 to 2048 characters). Optional: recheck (start a fresh check instead of reusing a result from the last 24 hours; allowed once the last check is an hour old) and wait_seconds (0 to 45, default 40; a check takes 20 to 60 seconds, and if it is still running you can poll with get_report). An EU browser visits the page twice so consent behaviour is exercised.

• get_report — read the result of an earlier check. Required: report (a report id such as Ab3dE5fG7hJ9kL2m, or the full report link, 8 to 200 characters). Returns the status, a grade per pixel found, each problem with its explanation and fix-guide link, and a checklist. Reports are kept for 90 days.

• list_findings — list every problem a check can report. Takes no arguments. Each entry carries its id, severity, title and explanation, and whether a fix guide exists for it. Useful for learning the catalogue before you run a check.

• get_fix_guide — read the step-by-step fix guide (Markdown) for a problem. All arguments are optional: issue_id (a problem id from a report, for example meta.fires_before_consent), platform (for example framer, shopify, webflow, wordpress; when omitted or unknown the general version is returned) and consent_tool (one of complianz, cookiebot, cookieyes, onetrust, usercentrics; when given, the guide covers wiring the Meta, TikTok and OpenAI pixels to that consent tool). Setup guides are also available through issue_id, for example the Meta Conversions API, the TikTok Events API, and installing the OpenAI Pixel or its Conversions API.

Related MCP server: AI Readiness

Configuration

• ISPIXELWORKING_MCP_URL (optional) — override the upstream endpoint. Defaults to https://ispixelworking.com/mcp. (The override hook is this rebuild's own labelled reconstruction.)

• ISPIXELWORKING_API_KEY (optional) — when set, every upstream request carries an Authorization: Bearer <key> header. When it is not set, no key header is sent. (The header name is pinned by the original's registry record, which declares a secret Authorization header; the record pins no package env name, so the Bearer form and forwarding the key from this env var are labelled reconstructions — the original surface was captured keyless.)

Forwarding mechanics

On the first tool call the client opens a streamable HTTP session upstream: a JSON-RPC initialize request (protocolVersion 2025-03-26, clientInfo name pixelgrade), then a notifications/initialized notification, then the tools/call request carrying the caller's arguments unchanged. When the upstream answers initialize with an Mcp-Session-Id header, that value rides on every later request of the session, which is reused for the life of the process. Replies may arrive as plain JSON or as server-sent events; both are read.

Successful upstream results pass through verbatim, including any isError flag the upstream sets. Failures — a JSON-RPC error, a non-2xx status, an unreadable response, or a request that cannot be completed — come back as an isError result whose text starts with pixelgrade error: (this rendering is this rebuild's own labelled reconstruction).

CLI

Each tool is also a subcommand taking one JSON argument string:

npx @reneturcios/pixelgrade check_page '{"url":"https://example.com/landing"}'
npx @reneturcios/pixelgrade get_report '{"report":"Ab3dE5fG7hJ9kL2m"}'
npx @reneturcios/pixelgrade get_fix_guide '{"issue_id":"meta.fires_before_consent","platform":"shopify"}'
npx @reneturcios/pixelgrade list_findings '{}'
npx @reneturcios/pixelgrade --version

Exit codes: 0 on success, 1 on an isError result, 2 on a usage error.

MCP server

npx -y @reneturcios/pixelgrade mcp

MCP client config:

{
  "mcpServers": {
    "pixelgrade": {
      "command": "npx",
      "args": ["-y", "@reneturcios/pixelgrade", "mcp"]
    }
  }
}

Docker

docker run -i ghcr.io/cloudcorprecords/pixelgrade
docker run --rm ghcr.io/cloudcorprecords/pixelgrade --version

The container's default command is mcp, so the first form is a stdio MCP server; any other arguments reach the CLI normally.

Development

npm install
npm run build
npm test
npm run dev

The test suite runs entirely against a stubbed JSON-RPC upstream — no live network calls.

License

MIT — see LICENSE.