SitePulsar MCP Server

What SitePulsar MCP Server does

The SitePulsar MCP Server MCP server connects an MCP client to SitePulsar's hosted Agent Experience Optimization platform. It evaluates websites across three pillars: FIND measures discoverability in major AI answer engines; READ covers machine comprehension, structured data, and trust signals; USE examines whether autonomous agents can take action on a site.

The service supports both quick checks and deeper investigations. A readiness check returns immediate signals without starting a full audit. A full audit produces an identifier that can be polled for completion, followed by an overall AEO score, pillar scores, key findings, the weakest area, and recommended fixes. Results can also include section-level scores, crawl and robots information, schema findings, rendered-DOM observations, reputation across engines, competitor grouping, and agent-oriented probes.

How it works

Connect an MCP-compatible client to https://mcp.sitepulsar.ai/mcp using streamable HTTP. The server is remote and hosted, so the repository does not need to be built or run locally. Once connected, the client can call the available tools and pass website URLs, brands, queries, or audit identifiers as appropriate.

For a detailed audit, call run_audit first and retain its returned audit_id. Use get_audit to check progress and retrieve the summary when the audit finishes. get_audit_detail, get_audit_full, and get_fixes expose progressively more detail or a prioritized action plan. compare_aeo supports comparisons across multiple URLs using the same pillar scale.

The additional probes inspect external signals rather than only the target site's page content. They can check where AI engines mention a brand, look for related servers and packages in the MCP registry, PyPI, and GitHub, inspect an advertised MCP endpoint, examine /.well-known/ucp, or extract structured product and offer information from a product page.

Setup and configuration

There is no local installation requirement. Add the hosted endpoint as a custom connector in Claude, or add the URL to an MCP configuration for Cursor or Cline. Claude Code can register it with an HTTP MCP command using the same endpoint.

Interactive clients can use OAuth. The protected-resource metadata is published at https://mcp.sitepulsar.ai/.well-known/oauth-protected-resource, allowing the client to discover the authorization flow. Headless or non-interactive clients can instead send Authorization: Bearer <mcp_api_key> on each request. API keys are created in the SitePulsar dashboard, and the documented key-based plan is Pro or Business.

The SitePulsar MCP Server MCP server does not require environment variables to connect. Plan and billing requirements apply to metered operations rather than to the transport itself.

Tools and capabilities

• check_agent_readiness: Performs a free, synchronous URL check for readability, structured data, and agent accessibility.
• run_audit: Starts a metered asynchronous AEO audit.
• get_audit: Polls an audit and returns its decision-oriented summary when ready.
• get_audit_detail and get_audit_full: Retrieve detailed findings, or combine summary, detail, and fixes; expand can reduce the full response.
• compare_aeo: Compares several URLs on the FIND, READ, and USE scale.
• get_fixes: Returns impact-ordered, pillar-labeled recommendations.
• search_companies and scan_visibility: Check brand or company visibility in AI engine results; visibility scans are metered.
• probe_agent_discovery, probe_mcp_functional, and probe_ucp_readiness: Inspect public discovery, MCP endpoint declarations, and UCP readiness signals.
• scan_product_page: Extracts structured product and offer information for agent readability; its pricing can vary by use.

Metered operations require a Pro or Business plan. Other tools are described as free in the provided tool list, subject to the stated exception for scan_product_page.