seo
Free Google Search Console MCP server and CLI for AI agents. Reads your real Search Console, Google Analytics, Bing Webmaster Tools, Clicky and Simple Analytics data, plus DataForSEO and Ahrefs keyword research. Use it for any SEO question - why Google traffic or clicks dropped, organic traffic decline, quick wins, low CTR, striking distance keywords, content decay and declining pages, keyword cannibalization, indexing and URL inspection, one page's queries, Google update impact, the...
npx skills add https://github.com/iannuttall/seo --skill seoseo
seo reads Search Console, Google Analytics, Bing Webmaster Tools, Clicky and
Simple Analytics data, crawls sites, checks live pages, and gets keyword and
backlink data from DataForSEO or Ahrefs.
Use the MCP tools when they are connected (hosted at
https://seoskill.dev/mcp, or local with seo mcp serve). Otherwise use the
CLI: seo, or npx -y seo. Add --json for JSON.
seo crawl --url <url> needs no sign-in and works on any site, local and
staging too. The other commands need the person's data: seo login prints a
link for them to approve, then run it again.
| CLI | MCP tool | Returns |
|---|---|---|
seo next --site <site> | seo_next | The next step in the SEO loop for the site, with its jobs; start here when asked to work on a site's SEO |
seo report --site <site> | seo_report | Search Console overview and sections; with no site, last week across the sites in the weekly site audit; --section teardown when asked to write a teardown of any site |
seo page --url <url> | seo_page | One URL: live page, index status, queries |
seo compare --site <site> --date <date> | seo_compare | Before and after a date, with Google updates |
seo index --site <site> | seo_index | Google's indexing reasons, labelled, and pages that lost their clicks |
seo submit <url...>, --indexmysite | seo_submit | Ask Bing and others to crawl pages (IndexNow, free, after seo submit key <site>), or send them to Google with Index My Site (the person's credits; quote first, send with --yes or confirm: true only after they agree) |
seo research <action> | seo_research | Keywords, keyword ideas, SERP (organic position, mobile) and SERP overlap, whether Google indexed any URL (index_check), top pages, competitors, backlinks, pages by backlinks, link gaps, AI answer citations, link and listicle targets (--pages, or ai_answers with competitors), for a site, subfolder (example.com/blog/) or page; provider optional; next pages are free for a day |
seo data --site <site> | seo_data | Raw Search Console, Bing, Clicky, Simple Analytics or Google Analytics rows |
seo crawl --url <url> | seo_crawl | Technical issues across a site |
seo internal-links --site <site> or --url <url> | seo_internal_links | Links to add across a site, or to one page |
seo lint <files> or --url <url> | seo_lint | Lines in a draft, title, description or live page that read as AI-written, each with a fix; --only ai for AI patterns alone; no sign-in |
seo feedback | seo_feedback | Send a bug or idea about these tools, with your proposed fix, to the maintainer; show the person the text and send only if they agree |
seo sites | seo_sites | The properties the person can read |
seo site new <folder>, then seo site guide <step> | CLI only | Build a new data-driven site on Astro and Cloudflare, one step at a time. Its build runs seo site check |
Each result ends with the command to run next. Start small and ask for more:
a site, then a --section (section), then --detail rows or full.
seo <command> --help shows the options.
The SEO loop
When the person asks you to use seo or improve a site's SEO without a
narrower question, run seo next. It keeps the site's business context
(what it sells, who buys, conversion pages) and its issues list, and returns one
step: research the business context, unblock pages, improve pages that rank,
or plan new pages and earn links. Do the step's jobs, ask the person before anything goes
live, mark each job, then run seo next again.
Issues kept between sessions
seo report --site <site> --section issues (section "issues") checks the
site again and keeps a list of what it found, each with an id, the latest
check and what the person marked it. Record what the person decided:
--mark fixed or --mark ignored with --issues <id>,<id> (mark,
issues, optional note). "Fixed" is the person's word; the list says
when the check still finds it. The hosted server, and a CLI linked with
seo login, keep the list with the account, for every site; an unlinked CLI
keeps it on this machine.
If the person has a list saved from before, such as .seo/issues.json,
pass its contents once as saved to carry its marks over.
Google updates
When the person asks whether a Google update hit the site, run
seo report --site <site> --section google-updates (section "google-updates"). It compares the latest ranking update's days with the
same days before and says whether that is outside the site's normal
movement. --end-date (end_date) picks an earlier update. Report it as
timing, never as the cause.
Indexing reasons each week
seo index inspects known URLs in a fixed order (lost-click pages, Search
Console pages by clicks, the sitemap, the rest), 100 by default, and labels
each Google reason fix, check, expected or indexed from a live check. In the
CLI or local MCP, --max-urls (max_urls) asks for more, up to today's
quota, and reuses saved inspections up to 7 days old, so paging or --detail full costs no quota. Run it once for a baseline and keep the full result (--detail full).
Run it again each week with the same number and compare by URL and reason: a
new reason, or a URL that moved to fix, is what to report.
seo keeps no history, so you store the results. Once this runs weekly, the
person can turn off Search Console's indexing emails. For one URL from such an
email, run seo page --url <url>.