MCP Server now online for AI-powered web audits

The most important points at a glance!

• The SEOFuxx MCP Server provides AI assistants with standardized tools for projects, audits, recommendations, and sitemaps, and authenticates via your personal API key.
• You start new page analyses, read the latest results and Core Web Vitals, and filter sitemap reports without a new crawl at the project level.
• The next extension will add Google Search Console features (check connection, list properties, submit sitemap) for an end-to-end SEO workflow.
• The getting-started steps show you setup, authentication, tool calls, and an example request-response flow for ChatGPT or Claude.

MCP features

Through the SEOFuxx MCP Server, your AI assistant can access the projects and analyses in your SEOFuxx account. You connect it with your personal API key from your account. The functions are provided as tools via the same server.

Projects and SEO audits

• Show projects: list_seo_projects
  returns your saved projects with project ID, website URL, keywords, and auto-crawl status. This helps your assistant first find the right project in your account.
• Fetch SEO recommendations: get_seo_recommendations
  reads the most recently saved analysis of a URL. You get evaluations, existing PageSpeed and Core Web Vitals data, plus prioritized recommendations with guidance and code examples for implementation.
• Start a new page analysis: run_page_audit
  runs a new SEO audit for a URL. Optionally, you can specify a focus keyword. PageSpeed and Core Web Vitals are then determined in the background; your assistant can retrieve the supplemented results later.
• Fetch sitemap report: get_sitemap_report
  reads your account’s sitemap report: key figures from the last crawl (total URLs, OK, redirects, 4xx/5xx errors, URLs no longer included in the sitemap), the latest crawl runs, and the problem URLs with HTTP status and error. Optionally, you narrow the report to a domain with domain . No new crawl is started; the report shows the most recently saved state.

Google Search Console and sitemap submission

With the next server update, features for Google Search Console will be added. This lets you check your Google connection, select suitable properties, and submit the public URL of your sitemap directly to Google.

• Check Google connection: get_search_console_connection shows whether Google Search Console is connected to your SEOFuxx account and whether permission to submit sitemaps is in place.
• Show Google properties: list_search_console_properties returns the available Search Console properties and their permissions. You can narrow the selection to your project using the project ID. A project can have multiple suitable properties, such as a domain property and URL-prefix properties for individual areas of your website.
• Submit sitemap to Google: submit_sitemap submits the public URL of your sitemap or sitemap index to Google. Your assistant uses an existing project from your account and a suitable Google property with write permissions.

Before submitting a sitemap, your website must exist as a project in your SEOFuxx account. Also connect Google Search Console under Account → Search Console and grant the required permissions. Your XML sitemap must already be publicly accessible on your project’s host. The Google link remains usable for MCP even after logging out of the website; you can end this access via “Disconnect” in your account.

An example request to your AI assistant:

“Check whether seofuxx.com exists as a project in my account, and show me the suitable Google properties. Then submit https://www.seofuxx.com/sitemap.xml to Google.”

If multiple properties fit, your assistant asks for your selection. A successful submission confirms that Google has accepted the sitemap URL. Google then decides on crawling and indexing of the included pages.

Introduction

The SEOFuxx MCP Server serves as a mediating layer between your AI assistant and the SEOFuxx API, so you can retrieve audits, recommendations, and sitemaps in a standardized way via tools. With uniform schemas for input and output, you reduce prompt glue, minimize miscalls, and increase the reproducibility of your analyses. For production-like workflows, it’s especially relevant that requests are designed to be idempotent, rate limits are clearly communicated, and results are unambiguously referencable with timestamps, project IDs, and resource IDs. This way, you integrate the MCP Server into ChatOps playbooks, CI pipelines, or recurring SEO reviews with clear SLAs and get measurably consistent results.

Basics: architecture and authentication

The MCP Server acts as a tool provider for your AI client: you authenticate with a personal API key that checks permissions at the project level and makes all access auditable. Requests include a stable project reference that deterministically assigns tool calls like run_page_audit or get_seo_recommendations to a context. Responses include metadata (time, source, measurement window, analyzer version used) so you can compare, cache, and repeat results if needed. Errors are returned with unique codes (for example, INVALID_PROJECT, RATE_LIMIT, ANALYZER_BUSY) so your assistant can follow up specifically, wait, or choose alternative paths.

Context: use in AI assistants

In practice, you orchestrate tool calls via your assistant, which first uses list_seo_projects to identify the target project, and then contextually starts audits or reads recommendations. Clearly defined parameters like target URL, optional focus keyword, and domain filter for sitemaps let you control the scope and cost of analysis. At the same time, structured results with scores, evidence links, and prioritized recommendations enable automatic preparation into checklists, Jira tickets, or editorial briefs. This ensures the bridge between analysis and implementation is robust and works without copy-paste losses.

Problem areas: typical sources of error

Integrations often fail due to ambiguous project assignments, missing permissions, or incorrectly set rate limits, leading to inconsistent results or timeouts. Mixing old and new analyses without time context also distorts trends when assistants combine results from different analyzer versions. Also avoid spontaneous parallel audits of the same URL with identical parameters if you haven’t implemented deduplication, as this creates unnecessary load and consumes quotas. Plan backoffs on ANALYZER_BUSY and check before writing whether resources (project, property, sitemap URL) exist and are associated with your key.

Causes: why integrations become unstable

Instability often arises from missing retries with jitter, unclear timeout budgets, and the absence of correlation IDs, which makes incident diagnosis difficult. Equally problematic are unvalidated inputs like non-canonical URLs, which lead to duplicate entries and divergent measurement points. If your assistant parses unstructured responses instead of using defined fields, the error rate rises in ticket creation or reporting. Clean URL normalization, consistent encoding, and carrying over the provided IDs into downstream systems are therefore basic prerequisites for robust operation.

Verification and tools: observability and security

Rely on request logging with correlation IDs, count tool calls per project, and measure success rates, latencies, and error types separated by tool. For production environments, enable limit alerts and run a warmup call after deploy to prime caches and analyzer containers. Rotating API keys, scopes per environment (dev, staging, prod), and read-only keys for reporting reduce the risk of misuse. Add a confirm prompt in the assistant for write operations like submit_sitemap and validate host matches between project and sitemap URL in advance.

Getting started

Start with a personal API key from your account and store it in your AI client as a secure secret parameter. Call list_seo_projects to pick your target based on the project ID and start an analysis for the canonical URL with run_page_audit , optionally with a focus keyword. Wait for the first finding and later fetch extended metrics via get_seo_recommendations once PageSpeed/Core Web Vitals have caught up. For sitemap health, use get_sitemap_report with an optional domainfilter; plan a polling interval of 15 to 30 minutes for dashboards and log the response IDs for traceability.

Conclusion

With the SEOFuxx MCP Server, you integrate robust SEO workflows into AI assistants, reduce errors through clear tool interfaces, and speed up your path from measurement to implementation. If you consistently adhere to authentication, project context, rate limits, and observability, you achieve reproducible analyses and a higher implementation speed in the team. Gradually extend your setup with Search Console tools to seamlessly close the cycle from crawling to diagnosis to submission and make your visibility more predictable.