Cursor's agent already writes your code. Give it a real technical SEO auditor and it can crawl staging before you deploy, group what it finds by template, fix the templates in the files you have open, then crawl again and show you what changed. bseoa runs on your machine as an MCP server, so it reaches localhost and private staging hosts as well as the live site.
14 days, every feature, no page limit, no credit card.
The local server is the full product: JavaScript rendering, no page budget, and crawls that run from your own machine. Setup takes a few minutes.
Download the bseoa CLI for Windows, macOS or Linux and put it on your PATH. Start the free trial to get a license key.
Once, and every host on this machine can use it:
bseoa license store <key> bseoa license status
Put the block below in ~/.cursor/mcp.json for every project, or in .cursor/mcp.json in one project. Then open an Agent chat and ask.
{
"mcpServers": {
"bseoa": {
"command": "bseoa",
"args": ["mcp"]
}
}
}
bseoa and black-seo-analyzer are the same binary; if yours is installed under the long name, use that name in "command" or add a symlink. Cursor does not always inherit your shell PATH: if the server will not start, put the absolute path from which bseoa (where bseoa on Windows) in "command". Skipped step 2? Add "env": {"BLACK_SEO_ANALYZER_KEY": "<key>"} next to "args". The license is read when the server starts, so restart Cursor after storing a key.
Cursor's agent reads Agent Skills. Download the bseoa skill and unzip it so that SKILL.md ends up at:
~/.cursor/skills/bseoa/SKILL.md
Cursor also loads skills from ~/.agents/skills and ~/.claude/skills, so one copy can serve several tools. Type /bseoa in Agent chat followed by what you want. The skill teaches the agent the CLI, including how to install it when it is missing. Details are on the skill page.
Cursor also takes remote MCP servers by URL and supports OAuth sign-in for them. Use this block instead, and sign in with your email when our page opens:
{
"mcpServers": {
"bseoa": {
"url": "https://mcp.blackseoanalyzer.com/mcp"
}
}
}
Crawls run on our servers and the results stay in your account, shared with Claude.ai and ChatGPT. No JavaScript rendering, and no private or localhost addresses.
These work as written in Agent chat. demo.blackseoanalyzer.com is a made-up bakery site we keep with known problems; it stands in for your staging host and your preview URL until you swap in your own.
Audit staging before you deploy
Crawl https://demo.blackseoanalyzer.com/ up to 100 pages as if it were our staging site. List every high or critical issue that should block the deploy.
The agent runs start_crawl, waits on crawl_status, then reads issues_summary and query_issues filtered by severity. Locally, the crawl can reach http://localhost:3000 or a staging host on your private network.
Group issues by template and write the tickets
Group the issues from my last crawl by template. For each group write a ticket with the issue key, the fix and the affected URLs, and name the file in this project that renders that template if you can find it.
One template usually produces hundreds of rows of the same issue key. The agent collapses them into a handful of tickets, and because it can search your codebase, it can point at the layout or partial behind each one.
Fix, crawl again, compare
Crawl https://demo.blackseoanalyzer.com/ again, then compare it with the previous crawl. What did my changes fix, and what is new?
The agent finds both sessions with list_crawls and runs compare_crawls: issues fixed, issues new and issues still there. Only the new ones are regressions.
Check a pull request preview
Check https://demo.blackseoanalyzer.com/menu as if it were the preview URL for this pull request. Would anything on it be an SEO regression if it shipped?
check_page audits one URL in seconds without following links. Paste your preview URL in its place. If your preview host marks previews noindex on purpose, tell the agent to ignore that one.
Write the robots.txt change
Read https://demo.blackseoanalyzer.com/robots.txt and write the change that keeps crawlers out of /contact while everything else stays crawlable. Show it as a diff, then tell me whether Disallow or noindex is the right tool for that page.
The demo's contact page is already noindex, which makes this a trap worth knowing: a Disallow stops search engines fetching the page, so they never see its noindex. Test the result with the free robots.txt tester.
Write the client report
Write a client report from my last two crawls of demo.blackseoanalyzer.com: what was fixed, what is still open, worst first, with the affected URLs. Save it as seo-report.md.
The agent reads compare_crawls and issues_summary and writes the file into your project, ready to edit and send.
bseoa reads and reports; it never logs in to your site or changes it. Cursor changes your code, and bseoa tells it whether the change worked.
Every feature, no page limit, no credit card. Covers the CLI, the desktop app and the local MCP server in Cursor.
Everything local, updates included. The Agency License is $1,999 one-time. Every purchase carries the source code guarantee.
With a license; $50/month or $500/year without. 25,000 crawled pages every rolling 30 days, 2,000 per crawl, results kept 90 days.
The local server needs no hosted plan. The hosted trial is separate: 14 days and 5,000 pages from your first hosted sign-in, results kept 30 days. Single-page checks never count against a hosted budget. Compare plans on the pricing page.
crawl.db in the directory the server starts in. To keep one history across projects, add "--db-path", "/absolute/path/crawl.db" to "args" after "mcp".A few lines of mcp.json, then ask. 14 days free, no card.
Start the free 14-day trialUsing another tool? Black SEO Analyzer for Claude Code and for Codex. Every host is on the connector page, and how bseoa compares with other SEO MCP servers is on SEO MCP servers compared.