The Search Console MCP server with full read and write access

Google Search Console MCP connects Search Console to AI assistants through the Model Context Protocol. Ask Claude or ChatGPT for query and page reports, inspect URL indexing, review sitemaps, and prepare approval-gated sitemap submissions or removals in the same conversation.

No. As of August 2026, Google does not publish an official Search Console MCP server. The options are community open-source servers, building your own on the official Search Console API, or using a managed read and write layer like Markifact.

Open-source servers are free but self-hosted: you create Google API credentials, run the server, and get Search Console only. Markifact is hosted, connects with one OAuth flow, supports multiple sites, and puts Search Console in the same conversation as Google Ads, GA4, Meta Ads, TikTok Ads, Sheets, and Slack.

Markifact cannot edit the content or code on your website, but Search Console MCP now has read and write access for supported Search Console management actions. It can submit or reprocess a sitemap and remove an existing sitemap submission after you approve the exact operation.

The gsc_get_report operation covers clicks, impressions, CTR, and average position by query, page, country, device, and date. gsc_list_report_fields discovers the schema, gsc_select_accounts switches properties, gsc_inspect_url checks indexing, and the sitemap tools list, submit, or remove feeds.

Yes, and it is one of the best uses. Ask for queries at position 8 to 15 with meaningful impressions and the AI pulls the report, ranks the opportunities, and can draft the content plan in the same conversation.

Yes. Compare two periods, find the pages losing clicks, and let the AI separate position losses (a rankings problem) from impression losses (a demand problem), which point at different fixes.

Yes. This is where a multi-platform MCP earns its keep: find queries you rank first for organically while still paying for the same clicks in Google Ads, or queries worth defending with paid because organic slipped. Both reports run in one conversation.

Yes. Connect all your verified Search Console properties and switch between them mid-conversation. Agencies can answer the same question across every client site without reconnecting.

No. Running a Search Console MCP yourself means creating Google Cloud credentials and managing tokens. With Markifact you sign in with Google OAuth once and the hosted MCP server handles authentication.

Markifact uses a secure OAuth-based Search Console connection. Your tokens are never exposed to the AI client, reads run without approval, and every sitemap submission or removal stops for explicit approval before execution.

Yes for URL inspection and sitemap management. Inspect up to 100 URLs per run for index status, canonicals, crawl details, mobile usability, and rich results. You can also list, submit, reprocess, or remove sitemaps. Search Console does not provide a general API for requesting indexing of arbitrary pages, so Markifact does not claim to offer one.

Yes. Sheets and Slack are Markifact operations too, so a striking-distance keyword list can land in a spreadsheet, and a monthly organic summary can post to your team's channel, all from the same MCP session.

Yes. Google Trends MCP shows broader relative search demand, while Search Console shows the clicks, impressions, CTR, and rankings your own site received. Comparing both helps separate a ranking problem from a market-wide change in demand.

Yes. Google Search MCP retrieves the current organic result set for a query, while Search Console shows how your own pages performed over time. Use both to compare live competitors with your clicks, impressions, CTR, and average positions.

Yes. Markifact works as a Google Search Console connector for Claude. A connector is Claude's name for an MCP server in its interface: add Markifact under Settings, then Connectors, enable the Search Console tools, and Claude can work with Google Search Console directly in the chat. The same connector also works in ChatGPT and other MCP clients.