Grok

How the integration works

Grok supports custom MCP connectors: any Model Context Protocol server reachable over the public internet can be attached and its tools used in conversation. GSC Wizard is exactly that — a hosted MCP server at https://mcp.gscwizard.com/mcp — so no new Search Console integration is involved. Grok reuses the same endpoint, the same Google authorization, and the same warehouse that Claude and ChatGPT use. There is no Search Console tile in Grok’s built-in connector catalog, so the custom connector is the route today — and custom connectors are available on every Grok plan, including the trial, so a failed connection is the endpoint or the sign-in, never the subscription.

Whichever you use, the analysis runs server-side. Ranking changes, decay, cannibalization, CTR curves, and opportunity scores are computed against the full dataset before Grok sees them, so the model interprets a finished result instead of doing arithmetic over raw rows.

Adding the connector

1. In GSC Wizard, create a key under Account → API keys. Pick read only unless you want Grok to be able to write.
2. Open grok.com/connectors, choose New Connector, then Custom.
3. Enter https://mcp.gscwizard.com/mcp as the MCP server URL.
4. Complete the authentication step, then ask about a property in plain language.

Grok discovers the server’s tools on connect. Start with “show me the Search Console properties I can access” to confirm the account behind the connector is the one you expect.

Authentication and scopes

GSC Wizard accepts two authentication modes, and MCP clients use whichever they support.

What Grok can do once connected

The server exposes over a hundred tools, but most SEO questions resolve to a handful. Naming the analysis you want in the prompt improves tool selection considerably.

Date ranges are optional on every tool. Omit them and the server uses the last 28 settled days, accounting for the reporting lag, with the comparison window defaulting to the previous period of the same length — so Grok never has to know today’s date to ask a sensible question. The full catalog is on the GSC MCP page.

The SEO analyst Bot

A Grok Bot is a persistent agent with a fixed job, its own instructions, and its own routines. Attaching the GSC Wizard connector to a Bot keeps the rules that make SEO analysis trustworthy in place instead of restating them every session.

The instructions worth setting are the ones that name how models get SEO data wrong: prefer the server-side analysis tools over raw rows, always state the analysis and comparison periods, keep facts separate from interpretation and recommendations, do not read correlation as causation, prioritize by impact rather than by position movement, treat a lower numerical position as better, and never call a write tool without explicit intent.

Routines turn a proven workflow into a schedule: a daily exception watch that stays silent unless a threshold is crossed, a weekly report, a monthly executive summary, and a monthly “did last month’s changes work?” check against your annotations. Add one only after the same analysis has held up interactively — a scheduled report that quietly says the wrong thing is worse than no report.

Grok Build

Grok Build, xAI’s terminal coding agent, supports remote MCP servers. Registering GSC Wizard there lets the agent read the performance of the pages it is about to change.

{
  "mcpServers": {
    "gsc-wizard": {
      "type": "http",
      "url": "https://mcp.gscwizard.com/mcp"
    }
  }
}

Grok Build runs a browser sign-in flow for MCP servers that use OAuth, so the key never has to touch the config file. If you use a static header instead, keep the key in an environment variable and out of source control. Verify the exact syntax against your installed version; Grok Build is moving quickly.

Plugins bundle an MCP server with skills, slash commands, and agents, so a whole analysis runs from one command: a weekly report, ranking changes, decay, cannibalization, or opportunity scoring.

The xAI API

Remote MCP tools are supported in xAI’s native SDK and in its OpenAI-compatible Responses API, so an agent you build can call GSC Wizard tools directly by passing the server URL and an auth header.

Note that require_approval and connector_id from OpenAI’s version of the parameter are not supported on xAI — gate write access with a read-only key rather than an approval callback.

If you would rather not speak MCP at all, the same keys, scopes, and rate limits back a REST surface at https://mcp.gscwizard.com/v1, with a generated OpenAPI 3.1 document at /v1/openapi.json that works as a function-calling spec. See the code example.

Troubleshooting