SEOLint alternative: what to use now that it has shut down

SEOLint shut down in July 2026. Its site now opens with "SEOLint is no longer operating" and keeps the pages as an archive. To replace it, you need four things: an MCP server your agent can ask what to fix, a scheduled scan, a regression check on pull requests or previews, and priorities based on your Search Console data. No single free tool does all four. The usual replacements are a free Search Console MCP server for your real queries, a free Claude Code skill such as claude-seo for on-demand audits, a paid data MCP (Ahrefs, Semrush or SE Ranking) if you need market data, or Serpkin, which covers the loop of next fix, preview check, weekly pull request and measured result.

We make Serpkin, so weigh that. Everything about SEOLint below comes from its own archived pages, and every other tool's details come from its official docs or pricing page, checked on November 3, 2026 and listed under sources.

#What happened to SEOLint

SEOLint was an SEO agent for developer-led sites. Its archived homepage describes a weekly site report, a ranked to-do list "sorted by Search Console traffic impact, not severity scores", paste-ready fix prompts for Claude Code, Cursor and Windsurf, pull requests for mechanical fixes (titles, meta, alt text, schema, sitemap), blog drafts, and GitHub Actions that failed builds when fixed issues came back. Its MCP server (the seolint-mcp npm package) let an agent scan a URL, list open issues, read site history and ask for the next action.

The founder's post-mortem, dated July 2026, sums it up in one line: "The agent worked. The business did not follow." It is an unusually honest write-up and worth reading if you build developer tools. The technical loop ran in production. What didn't work was distribution, and the fact that the product's best feature (grading each fix against Search Console at 14, 28 and 90 days) landed after a 7-day trial ended.

#What you need to replace

Split SEOLint into the jobs it did for you. Most people relied on two or three of them, not all six. Replace the ones you actually used.

#How to migrate off SEOLint

Do this even if you don't replace it yet. A dead MCP server can slow down or confuse every agent session, and old credentials should not sit in your configs.

1. Remove the MCP server from your agent

   SEOLint's docs installed it as a server named seolint. In Claude Code, list your servers and remove it:

   claude mcp list
   claude mcp remove seolint
   
   # Find any other config that still mentions it
   grep -rn "seolint" .mcp.json .cursor/mcp.json ~/.cursor/mcp.json ~/.codeium/windsurf/mcp_config.json 2>/dev/null

   Delete the seolint entry from any file the grep finds, then fully restart the client. For Claude Desktop, edit claude_desktop_config.json the same way.

2. Delete the API key

   Remove SEOLINT_API_KEY and SEOLINT_API_URL from your MCP configs, shell profile, CI secrets and any .env file. The key no longer works, but leftover secrets make audits noisy.

3. Remove CI steps and the GitHub App

   Search your workflows for SEOLint and delete those steps, or the job fails once the API is gone. SEOLint opened pull requests through a registered GitHub App: uninstall it under your GitHub settings, in Applications, Installed GitHub Apps (or the organization's settings).

   grep -rn -i "seolint" .github/ package.json
   gh secret list | grep -i seolint
   gh secret delete SEOLINT_API_KEY

4. Revoke its Search Console access

   SEOLint connected to Search Console with Google OAuth. Open your Google Account's third-party connections page and remove its access. If you added a user to the Search Console property, remove that under Settings, Users and permissions.

5. Keep what you can

   Your merged SEOLint pull requests are in git history. List them so you know which titles, meta and schema changes came from it, and compare those pages in Search Console (which keeps 16 months of data).

   git log --oneline -i --grep="seolint" --grep="fix(seo)"
   gh pr list --state merged --search "seolint" --limit 50

6. Add the replacement

   Pick from the options below and add one MCP server at a time, so you can tell which one the agent is calling.

#The options

Search Console MCP servers (free)

Google doesn't publish an official Search Console MCP server, but open-source ones work well. mcp-gsc (MIT) covers search analytics with period comparisons, single and batch URL inspection, and sitemaps. Destructive actions such as deleting a sitemap are off unless you set GSC_ALLOW_DESTRUCTIVE. It runs locally with uvx and your own OAuth client or service account, so your data goes from Google to your machine. This replaces SEOLint's Search Console priorities if you are happy to ask the agent to rank pages yourself (for example, "pages with high impressions and CTR under 2%"). It doesn't scan, schedule or open pull requests.

claude-seo and other free Claude Code skills

AgriciDaniel/claude-seo is a free, MIT-licensed Claude Code plugin with 26 sub-skills, 19 sub-agents and commands such as /seo audit, /seo technical, /seo schema and /seo geo. Its /seo google setup connects Search Console, PageSpeed Insights, CrUX and GA4 with keys kept on your machine, and /seo drift compares snapshots over time. Install it inside Claude Code:

/plugin marketplace add AgriciDaniel/claude-seo
/plugin install claude-seo@agricidaniel-claude-seo
/seo setup

It is the closest free match to SEOLint's fix prompts, because the agent that audits is the one that edits. The gap is everything that happens when you aren't in a session: the README describes no automatic schedule, no pull requests opened for you and no hosted history. You can run it on a schedule with your own CI job, but you maintain that.

SE Ranking MCP

SE Ranking runs a hosted server at api.seranking.com/mcp, with OAuth 2.1 or an API key header. It covers keyword research, backlinks, domain analysis and SERPs, plus your SE Ranking projects (rank tracking, audits, competitors), local data and AI visibility, though not every API endpoint is exposed. The data you get depends on your subscription, and you need API access on your account. Its project audits give you a scheduled scan, but the agent still turns findings into changes and nobody checks your previews.

claude mcp add --transport http se-ranking https://api.seranking.com/mcp
# then run /mcp in Claude Code to sign in

Ahrefs and Semrush MCP

If you already pay for a suite, its MCP server is the easy add. Ahrefs lists API and MCP access on Lite ($129/month) and up. Semrush MCP is read-only and comes with 50,000 MCP API units on Semrush One Starter ($199/month), Pro+ and SEO Classic Pro and Guru. Both are research servers: great for keyword and competitor questions mid-task, but they don't open pull requests or check your previews. Setup and metering for these and seven other servers are in SEO MCP servers compared.

claude mcp add ahrefs https://api.ahrefs.com/mcp/mcp -t http
claude mcp add semrush https://mcp.semrush.com/v2/mcp -t http

Serpkin

Serpkin is the option closest to SEOLint's shape. Its MCP server has 30 tools on every plan and the trial. serpkin_next_mission returns one ranked fix with target pages, source files, steps and acceptance checks, ranked with your Search Console data once connected. npx serpkin check <preview-url> fails on regressions before merge. npx serpkin weekly adds a Monday GitHub Action that has Claude Code implement the next mission and opens one draft pull request for you to review. After deploy, the mission is verified on production and results are measured at 14 and 28 days. Site watch alerts and the weekly summary can go to Slack or Discord. Differences: Serpkin doesn't draft blog posts, and it measures at 14 and 28 days, not 90.

#Side by side

#A free regression check for your previews

If the GitHub Action was the part of SEOLint you miss most, you can cover the basics with a short script. It fetches the same page from production and from a preview and fails when the preview loses a title, a canonical or gains a noindex in the page's robots meta tag.

#!/usr/bin/env bash
# Usage: scripts/seo-diff.sh https://example.com https://my-preview.vercel.app /pricing
set -euo pipefail
prod="$1$3"; prev="$2$3"
grab() { curl -sL "$1" | tr -d '\n'; }
has() { grep -qiE "$2" <<<"$1"; }
p=$(grab "$prod"); v=$(grab "$prev"); fail=0
for check in "<title>[^<]+" 'rel="?canonical' ; do
  if has "$p" "$check" && ! has "$v" "$check"; then echo "REGRESSION: $check missing on $prev"; fail=1; fi
done
if has "$v" '<meta[^>]+name="?robots"?[^>]+noindex' && ! has "$p" '<meta[^>]+name="?robots"?[^>]+noindex'; then
  echo "REGRESSION: noindex added on $prev"; fail=1
fi
exit $fail

Run it in a workflow on deployment_status events (Vercel sends one when a preview finishes) for a handful of key paths. It won't catch sitemap or robots.txt problems, and Vercel previews add an X-Robots-Tag: noindex header, so check the meta tag, not headers. npx serpkin check does the same comparison for up to 25 paths plus robots.txt, sitemap, H1 and description, and reports problems already on production as warnings instead of failures. The GitHub Actions guide has the workflow.

#How to choose

#Where Serpkin fits

SEOLint's post-mortem argued that the loop back to Search Console was the point: ship a change, wait, and find out whether it moved anything. We agree, and that loop is what Serpkin is built around. Setup is one command in your repository: npx serpkin connects Claude Code, Cursor or Codex, registers the site and shows the first mission. Both plans start with a 7-day free trial; Builder is $29 a month (3 sites) and Pro $79 (15 sites). It doesn't merge, deploy or promise rankings. See getting started, or what Claude Code can do on its own if you want to try the free route first. Ready to replace SEOLint? Start a 7-day free trial.