You can now use my list of Google algorithm updates and changes to AI Search via MCP so you can access it via any AI system that offers MCP access. Click here to expand the step-by-step setup guides to connect my MCP server to Claude, Antigravity, or wherever you use MCP.
I had this idea to turn my algo update list into an MCP. I rambled about it to Antigravity and used the /goal command and we completely rebuilt my algo update page and created this MCP.
Every update in my archive is maintained as an independent, transparent Markdown document with structured YAML frontmatter (verified date, category, platform, impact level, and reference links). This deterministic structure lets AI agents retrieve exact chronological facts without hallucinations.
Connecting and querying this MCP server is completely free. There are no API keys, paywalls, or fees on Marie's end. The tokens used to retrieve updates and converse with your agent are simply covered under your existing AI subscription (Claude Pro/Team, Antigravity, Cursor, etc.).
Once connected, your agent can access Marie's updates in natural conversation:
When diagnosing a drop or asking about an update, Claude or ChatGPT accesses my verified timeline, but will also often add its own SEO tips. My MCP server instructs the AI to separate them, but you can get crystal-clear formatting in your chat simply by asking:
Google released Gemini 3.8 Flash, making it available as a selectable option in the AI Mode model dropdown for paid subscribers. A new model option in AI Mode functions much like an algorithm update: the model's advanced reasoning allows it to directly resolve complex queries with fewer clicks to external sites, while simultaneously recalibrating which websites it predicts will provide the most valuable deep-dive source material.
Google published major updates to the Universal Commerce Protocol (UCP) specification, refining how e-commerce sites declare inventory, checkout workflows, and machine-readable product schemas for agentic shopping.
Google introduced direct controls in Search Console for Generative AI visibility. Webmasters can now monitor and set explicit preferences for how their content is summarized and utilized across Google's generative search features and AI Overviews.
Everything you need to know about privacy, how this MCP works, and getting the most out of it.
No, absolutely not. I have zero access to your data, your prompts, your GA4 or Search Console metrics, or your private client conversations.
This MCP server functions as a read-only reference library. When your AI agent (in Claude Desktop, Antigravity, or Cursor) asks for algorithm updates during a specific date window, it only asks my server for the matching public update facts. All of your proprietary traffic data, audit notes, and agent reasoning remain entirely private within your own AI client session.
No, it is 100% free. There are no API keys, paywalls, or fees on my end to connect or query this server. Any AI tokens used to retrieve updates and analyze your site are simply covered under your own existing AI subscription (such as Claude Pro/Team, ChatGPT Plus/Pro, or Antigravity).
Standard LLMs have knowledge cutoffs and frequently hallucinate exact Google update dates, confuse rollout timelines (e.g., when an update started vs. finished rolling out), or miss unannounced ranking tremors entirely.
Connecting to my MCP server provides your agent with deterministic, verified facts directly from my 15-year archive (from the original Panda update in 2011 through today's Gemini 3.8 Flash model rollouts and spam updates), eliminating guesswork and hallucinations.
Take the date window where you noticed an impression or click decline in Google Search Console or GA4, and ask your agent:
Your agent will automatically call the get_updates_by_date_range tool, review all confirmed and unconfirmed updates in that exact window, and correlate them with your traffic patterns.
Any platform supporting the Model Context Protocol (MCP) or standard SSE streams can connect, including:
I update the database continuously as new Google core updates, spam updates, search quality system shifts, and AI search changes (like Gemini models and AI Mode updates) are confirmed or observed in the industry.
Yes, fully! This server natively implements both the modern official Streamable HTTP transport standard (at /mcp) and the legacy HTTP+SSE transport (at /sse and /messages).
Streamable HTTP is the modern MCP standard (SEP-2596), unifying communication into a clean single-endpoint streaming model. Because many existing tools (such as Claude Desktop configurations using supergateway --sse) still use SSE, both transports run simultaneously so your agents will always connect seamlessly.