Goose SEO: Your Search Console as a Goose Extension

Goose can read and write files, run commands and use extensions. With SEOmatic as one of them, it can turn your search data into reports, plans and fixes.

An example on the Acme Plumbing demo site. You ask Goose:

“Write me a weekly SEO report for Acme Plumbing as a markdown file, with what to fix first.”

Goose calls SEOmatic:

• seomatic gsc_performance: Sep 28 to Oct 4 vs Oct 5 to Oct 11 (7 days each)
• seomatic gsc_insights: CTR outliers, Oct 5 to Oct 11 (7 days)
• write reports/seo-2026-10-12.md

Goose answers:

Saved to reports/seo-2026-10-12.md. The short version, Oct 5 to Oct 11 vs Sep 28 to Oct 4 (7 days each):

• Clicks: 412, up from 371.
• Winner: /emergency-plumber, up 38 clicks after it reached position 3.2 for “emergency plumber near me”.
• Fix first: /water-heater-repair had 980 impressions but a 1.1% CTR. Its title never says “repair”.
• Watch: /drain-cleaning-cost slipped from 6.8 to 9.4.

Want me to draft a new title for /water-heater-repair? On a paid plan I can stage it in SEOmatic for your approval.

Also try:

• “Which pages has Google not indexed, and why?”
• “Audit the site and save the results as a checklist”
• “Which pages rank on page two?”
• “Do AI answers cite our pages?”

Goose is a general agent: it works with files, folders and commands, not just repos. That makes it a good fit for SEO work that ends in a document instead of a diff:

• A weekly report saved as markdown, with exact date ranges and the one fix to do first.
• A content plan built from the queries your pages already rank for on page two.
• A client folder per site, if you look after several, with each report in its own place.
• Code fixes when the site lives in a repo, reviewed as a normal diff.

You bring the questions; SEOmatic brings the numbers from your Search Console; Goose does the writing and filing.

Try these in a Goose session, on the desktop or in the CLI:

• “Write me a weekly SEO report as a markdown file, with what to fix first.”
• “Which keywords rank on page two, and what would move each page up?”
• “List the pages Google has not indexed, and why.”
• “Audit the site and save the results as a checklist in this folder.”
• “My organic clicks fell 30% this month. Which pages lost them?”
• “Which questions do AI assistants answer without citing us?”

• Search Console performance and insights: clicks, impressions, CTR and position, period comparisons, CTR outliers and cannibalization, with exact date ranges.
• Indexing checks: whether Google indexed a URL, and why not.
• Site audit and page scores: up to 30 pages per crawl, or a 0 to 100 score for one URL.
• AI visibility and citations: where ChatGPT, Claude and Google AI Overviews mention or cite you.
• Keyword and backlink research on paid plans, with one free keyword preview.
• Tasks and campaigns on paid plans: fixes for CMS sites, staged for your approval.

Every tool, with its parameters and cost, is in the SEO MCP server reference.

1. You ask Goose a question about your site.
2. Goose calls the SEOmatic tools it needs through the extension, such as a period comparison or an indexing check.
3. SEOmatic returns your real data and its own analysis, with exact date ranges.
4. Goose explains what it found and, if you ask, writes a report, a plan or a code change on your machine.
5. For CMS sites on a paid plan, Goose stages the fix in SEOmatic instead. You approve it, and SEOmatic applies it with a rollback point.

About a minute, with Goose installed and a model set up.

1. Start the extension setup. In a terminal, run goose configure and choose Add Extension, then Remote Extension (Streamable HTTP). In the desktop app, open the sidebar, then Extensions, then Add custom extension.
2. Enter SEOmatic's details. Name it seomatic and use the endpoint https://app.seomatic.ai/api/mcp. Keep the default timeout and skip custom headers.
3. Sign in to SEOmatic. Start a Goose session. The first time Goose reaches SEOmatic, it opens the sign-in in your browser. Sign in or create a free account and pick your site.
4. Check the extension. In the desktop app, check that seomatic is turned on under Extensions. In the CLI, ask Goose which SEOmatic tools it has.
5. Ask about your site. Ask: “How did my site do on Google over the last 28 days?” The answer should name the date range and your real numbers.

On Windows, the config file is %APPDATA%\goose\config.yaml. The desktop app and the CLI share it, so you only add SEOmatic once. If the sign-in does not open, turn the extension off and on again, or use the API key header instead.

Pick goose configure or the desktop app. Both write the same extension, and Goose opens the sign-in for you. The skills route works too: npx skills add Minh42/seomatic-mcp -a goose.

Goose is an open-source AI agent first built by Block and now under the Agentic AI Foundation. It comes as a desktop app and a CLI, and both read the same config, so SEOmatic set up in one shows up in the other.

• The model Goose uses is your choice, from a hosted provider or a local one. SEOmatic connects to Goose, so switching models needs no new setup.
• Other open-source agents, like OpenCode or Hermes Agent, have their own setup on their own pages.

This page covers Goose on the desktop and in the terminal.

Connecting is free, and so is the SEOmatic account behind it: no card. A free account answers 5 questions a month from your own data, plus 10 bonus questions once you connect Google Search Console. Each tool call counts as one question, so a big request can use a few.

Free covers the diagnosis: Search Console reports and insights, site audits of up to 30 pages, live page analysis, and your AI visibility results. Live market data, like keyword volumes, backlink profiles and competitor domains, is part of paid plans, with one free keyword preview per workspace.

Paid plans remove the question limit and add the treatment: agents stage fixes on your site (titles, descriptions, content updates, technical fixes, internal links) and run campaigns and fresh AI visibility scans. Every paid plan starts with a 14-day free trial. See pricing.

Your model provider bills Goose's own work separately. SEOmatic questions count only the SEOmatic tool calls, not the files Goose writes.

Goose works on your machine and in SEOmatic, and both keep you in charge. On your machine, Goose's own permission settings decide what it may do without asking, and you review code changes before they ship. In your CMS, Goose can only stage a fix in SEOmatic. A staged change is a proposal: it says what will change and why, and nothing touches your site until you approve it. The assistant records your decision, and it is told never to approve anything on its own judgment. Approved fixes start within seconds, the page's previous state is stored, and any applied change can be rolled back. Changes that could take a page out of Google, like a noindex, need an extra explicit confirmation.

Using another open-source agent? See OpenCode, Hermes Agent and OpenClaw. Prefer an agentic terminal? See Warp. Learn what powers the answers: page indexing checks, technical SEO fixes and meta tags. All agents: AI SEO agents in every assistant. Full reference: SEO MCP server docs.

Using another agent? See OpenCode, Hermes Agent, Warp, or every AI agent.