Sitejar

Scan a web page for accessibility, security, privacy, SEO, quality and deceptive design issues from your AI client, then get fix guidance your own model can apply to the code.

View on sitejar.app

Hosted MCP Server

npx add-mcp 'https://mcp.sitejar.app/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Install

Connect your client

Pick your client. The remote endpoint is the same everywhere: https://mcp.sitejar.app/mcp

Claude Code

claude mcp add --transport http sitejar https://mcp.sitejar.app/mcp

First connection opens a sign-in on sitejar.app. Your existing account, MFA and all, works as-is. For every client step by step, follow the complete MCP server setup guide.

Tools

What your AI agent can do

Five tools, each one a plain call your agent makes on its own.

  • scan_website

    Scan a single web page across six compliance dimensions and return an overall picture plus per-dimension detail. Single-page by design; for whole-site scans, point users to the dashboard.

  • get_scan_result

    Fetch the result of a scan you started, including status while it runs, so your agent can wait and then read the findings.

  • get_issues

    List the issues found, filtered by dimension, severity, or count, so your agent focuses on what matters first.

  • get_category_result

    Drill into one dimension (accessibility, security, privacy, deceptive design, quality, or SEO) for its score and issues.

  • get_fix_guidance

    Get deterministic, source-level fix guidance for an issue, which your own model applies. No AI spend on our side; your agent does the writing.

Every scan covers all six dimensions: Accessibility, Security, Privacy, Deceptive Design Risk, Quality, SEO.

How it works

Connect, scan, fix

  1. Step 1

    Connect

    Add one line to your MCP client and sign in on sitejar.app.
  2. Step 2

    Scan

    Ask your agent to scan a URL. It gets issues across all six dimensions.
  3. Step 3

    Fix

    Your agent reads our fix guidance and writes the change. We never spend AI budget; your model does the work.

Built by a compliance company, so we sweat the auth

Sign in happens on sitejar.app, with your existing account, MFA and all. The MCP server uses OAuth 2.1 with PKCE and issues opaque, revocable tokens, so access can be cut immediately. Scans are single-page and rate-limited per plan. The server never forwards your token anywhere and never triggers our own AI. Recent studies found most MCP servers get authorization wrong; we set out to be in the small share that get it right.

Pricing

Free to try

Every plan, including Free, gets MCP access. Start with five single-page scans a month, no card required. Hit the limit and your plan's monthly quota lifts it, the same quota your dashboard scans use.

Not ready to connect? Run the underlying checks free in your browser, no account needed:

Questions

Before you connect

Which clients work?

Any MCP client that supports remote Streamable HTTP: Claude Code, Claude Desktop, Cursor, and others. stdio-only clients connect through the mcp-remote bridge, using the "Other (stdio bridge)" config on this page.

How do I sign in?

On first connection your client opens sitejar.app. You approve access there, with your normal account and MFA. We never see a password from the client.

How is this different from a dashboard scan?

The MCP server scans one page at a time, ideal for checking a page mid-build. For a full-site crawl with history and exports, use the dashboard.

Where does my data go?

The server scans the URL you give it and returns findings to your agent. It does not forward your token anywhere and does not run our AI fix engine.