PageLens AI

What PageLens AI MCP server does

PageLens AI MCP server exposes PageLens website audit data to MCP-compatible clients. It is intended for assistants that need to understand what failed in a website review, why a finding matters, and which issues should be addressed first. PageLens scans can cover security, SEO, performance, accessibility, UX, content, design, conversion, and QA journey audits.

The server provides account and domain context, recent scan summaries, detailed scan information, findings, quick-win rankings, and agent-oriented report resources. Scan data may include a launch verdict, builder platform, launch timing, severity totals, executive summaries, persona reviews, and QA journey information.

How it works

The service is hosted at https://www.pagelensai.com/api/mcp and does not require a local process. Configure an MCP client with this streamable HTTP endpoint, then authorize access through the browser when the client connects. Authentication uses OAuth 2.0 rather than an API key.

The available authorization scopes determine access. read:scans covers listing and reading scan results, read:findings covers findings and quick wins, and write:feedback covers finding feedback and owner decisions. The whoami tool can confirm the PageLens account and scopes associated with the session.

Tools and capabilities

PageLens AI MCP server exposes tools for:

• Listing verified domains and their badge information.
• Filtering recent scans by status, domain, or date.
• Reading complete scan summaries and QA Audit details.
• Paging through findings by severity, category, persona, URL, or rule ID.
• Returning high-impact, lower-effort quick wins, with optional preset-based ranking.
• Reporting false positives and other finding problems with reasoning and evidence.
• Recording acknowledged risks, intentional tradeoffs, accepted risks, or deferred fixes.
• Clearing a previous finding decision while retaining the audit history.

Two resources provide alternate report formats: a Markdown report for richer agent context and a compact JSON summary. QA Audit resources can include journey replay data, blocked or risky paths, safe actions, confidence, authenticated-route context, and page-budget coverage.

Setup and configuration

Add the remote endpoint to the MCP client configuration. Cursor uses a url entry in mcp.json; Claude Desktop uses an HTTP server entry with the same URL; and Codex CLI uses a URL under an MCP server configuration. Other compatible clients can use the streamable HTTP endpoint directly.

No local package installation or environment variable is documented. On first connection, the client opens a browser authorization flow for the PageLens account. The client must support OAuth-based remote MCP connections and the required scopes for the operations an agent will perform.

Limitations and notes

The server reads and updates PageLens report context; it does not itself describe a website crawl command or create a scan in the documented MCP tools. Feedback requires a paragraph explaining the issue and a concrete evidence snippet. Owner decisions annotate a finding but do not hide it, alter the report, or change the PageLens score.

QA Audit output depends on the captured journey and safe same-origin page discovery. Reports can identify blocked steps, pages reviewed against the page budget, authentication-profile status, and reasons fewer pages were captured.