You: Fix the SEO on our pricing page and keep going until it scores 90 or more.
Assistant: scan_page 46/100 no meta description, no H1, title 14 characters
get_fix_pack title, description and canonical, written from the page
schema · generate Product JSON-LD, 2 values to fill in
(edits pricing.html)
scan_page · html 92/100 before deploying, all errors gone
scan_page · rescan 94/100 live, +48 since the first scan
Done. 46 to 94. Two prices in the JSON-LD still need your real values.
New
- Drafts and post updates in WordPress. The content autopilot now ends in wp-admin: new posts arrive as drafts, updates show what changes, one click to approve, undo any time. Plugin 1.1.
- Content in your site's own voice. Content autopilot:
contentmodesvoice,calendaranddraftplus thecontent_autopilotprompt: your agent writes, OnPage.dev learns how the site writes, plans a dated calendar, checks each draft for voice, unsourced figures and SEO before you approve, and learns from measured results which formats work on your site. - Approve fixes in WordPress, one click. Your agent sends the fixes to the free OnPage.dev WordPress plugin. You see old and new value per field, approve, and undo any time. Through Yoast SEO, Rank Math or its own tags. OnPage.dev never logs in.
- Ask about your traffic, no CSV exports. Finds the day traffic broke and the Google update that ran then, from Google's own status dashboard. Plus lost and split queries, the questions AI answers, and Google against Bing, from the Search Console, GA4 and Bing rows your assistant already reads.
- Checks every deploy preview. Holds a Vercel, Netlify or Cloudflare preview against production before you merge: lost titles and canonicals, noindex, broken pages. In chat or as a GitHub Action that comments on the pull request.
They bring the data. OnPage.dev closes the loop.
The SEO MCP servers from Ahrefs, Semrush, SE Ranking and DataForSEO are built for data: what ranks, who links, what people search. Then the real work starts. OnPage.dev does that part, inside the same chat, and uses their numbers to decide what to fix first.
An example of the two kinds of answer. Prices, plans and the full comparison: onpage.dev/compare. Built by an SEO specialist with 24 years of experience; the scoring method is public at onpage.dev/methodology.
The whole loop, inside your AI agent
Most SEO tools stop at a report. OnPage.dev gives an agent every step, and each step is a tool it can call:
Four weeks later...
You: Did the fixes on /pricing and /features work?
Assistant: (Search Console) clicks for 28 days before and after 8 September
measure · impact changed pages +70%, unchanged pages +3%
→ +66% lift; /features +357% after the stray noindex was removed
measure · impact subtracts the trend of the pages you did not change, so seasonality and Google updates are not counted as your result. It is honest about what it shows: a comparison of two periods, not proof.
From finding to done
Most SEO tools stop at a report. OnPage.dev hands the plan to the tools your team already works in, so fixes get assigned and shipped. share · export turns a page scan or a whole site audit into:
- Google Sheets: one row per finding with priority, page, issue, fix, code and status. Every export comes with a private CSV link, so
=IMPORTDATA("https://onpage.dev/export/….csv")loads it into a sheet with no connector at all. - Slack, Teams or Discord: a ready summary with the fixes that matter most, posted by the agent's Slack tool or straight to an incoming webhook.
- Linear, Jira, Asana, Trello or GitHub Issues: one task per issue, grouped across pages, with priority, labels and how to fix it.
- Notion or Confluence: a Markdown checklist.
You: Audit example.com, put every fix in a Google Sheet and post the summary in #seo.
Assistant: site_audit · start 25 pages, 41 findings
share · export sheet link, Slack message, 12 tasks
(Google Sheets) rows added to "SEO fixes"
(Slack) posted to #seo
OnPage.dev never logs in to Slack, Google or your task board: the agent sends the plan with the tools the user already connected, or you paste one formula. Scan results and audits end with a hint, so agents offer the export on their own.
Seen by every reader
Google reads your HTML. Visitors see one screen on a phone. Screen readers, and AI agents that browse for people, read the accessibility tree. OnPage.dev checks all three in a real browser.
layout · render loads the page on a phone, tablet and desktop and checks the first screen: is the H1 and the call to action visible without scrolling, does a cookie wall or pop-up cover it, are tap targets big enough, what shifts while loading. It loads the page again with JavaScript off to show what GPTBot and ClaudeBot miss, and checks whether the first screen delivers what the Google snippet promises. With screenshots.
It also checks design and readability, the details that make a page look professional and easy to read:
- Text with too little contrast, measured against its real background (WCAG AA: 4.5:1, or 3:1 for large text), with the colours to fix.
- The exact element that makes the page scroll sideways on a phone, and by how many pixels.
- Mixed button styles: how many corner styles and heights the buttons use.
- Uneven spacing between sections.
accessibility · screen_reader reads the page the way a screen reader announces it:
heading level 1, Running shoes for flat feet
link, Women
link ← announced without a name
button ← announced without a name
image, Trail shoe in grey mesh, side view
link, Read more ← vague, and weak anchor text
It flags links and buttons without a name, vague link text, form fields without a label, a missing main landmark, skipped headings and visible text hidden from readers. Accessibility is not a direct Google ranking factor, but alt text, link text, headings and text that is not hidden are exactly what search engines, AI answer engines and browsing agents read. Even carefully built sites miss this, including sites built with AI coding assistants. In a real scan of a well-built news site that passed every other check, including Lighthouse, the screen reader view still caught links announced as just "link" in nine different card layouts across the site: each thumbnail image linked to its article but gave a screen reader no name to read.
Have your own browser tool (Claude in Chrome, Playwright)? layout · probe gives a small measurement script and layout · analyze analyses its results: no daily limit, and it works on staging, localhost and pages behind a login.
Real-user speed
Lab tests guess. web_vitals · now shows what real Chrome visitors got over the last 28 days, from the Chrome UX Report: the field data Google uses for ranking.
You: Is our homepage fast enough for Google?
Assistant: web_vitals · now Phone: passed · Desktop: passed
LCP 1.4 s good 93% of visits good
INP 197 ms good 75% of visits good, 8% poor
CLS 0.00 good 95% of visits good
- LCP, INP and CLS, plus FCP and time to first byte, on phone and desktop, each rated good, needs improvement or poor.
- Pass or fail on the Core Web Vitals assessment, and the share of visits that were good, needs work or poor.
- Pages with too little traffic fall back to the whole site, and the tool says so.
- What to fix for every metric that is not good. The same card is in every scan on onpage.dev.
Uses the SEO data you already have
Already connected Google Search Console, GA4, Ahrefs or Semrush to your assistant? OnPage.dev tells the assistant to pull their numbers in, then prioritize_fixes ranks every fix by the traffic it can win. Nothing connected? It ranks by severity, as before. You do not set anything up.
You: Audit example.com and tell me what to fix first.
Assistant: site_audit · start 10 pages, average 81
(Ahrefs) top pages clicks, positions, keywords, referring domains
prioritize_fixes ranked by traffic impact
1. /pricing is noindex but gets 900 visits a month. Fix this first.
2. /results ranks #6 for "league table" (5,600 searches a month).
Add the term to the H1 to push into the top 3, ~330 extra clicks.
3. /guide shows up 20,000 times at position 3 but gets 0.2% clicks.
Rewrite the title and description, test them with check_snippet.
How it works: an MCP server cannot read other connectors, and should not. OnPage.dev's results and instructions ask the assistant to fetch the numbers from any SEO data tool it has, and the assistant passes them to prioritize_fixes. The numbers are used for that answer only and never stored. OnPage.dev never logs in to those services.
Spikes, drops and the Google update behind them
Search Console, GA4 and Bing in one chat, no CSV exports. Ask about your traffic and the agent passes the daily rows to find_opportunities with mode: "anomalies":
You: Did anything odd happen to our traffic this quarter?
Assistant: (Search Console) daily clicks, 90 days, by page
find_opportunities · anomalies
Drop 25 to 27 Sep: -59% (90 clicks, normal about 220)
Google: September 2026 spam update (Ranking, 24 Sep to 8 Oct)
/pricing 12 clicks, normal about 80
- Each day is compared with the 28 days before it, per weekday, so a normal weekend dip is not an alarm.
- Google core updates, spam updates and Search incidents are read live from the Google Search Status Dashboard.
traffic_droptakes adrop_dateand checks the same list. mode: "ai_queries"finds the questions, comparisons and long conversational queries in your Search Console or Bing rows, per page, and flags pages that rank but lose the click to an AI answer.
And what AI already cites
With Ahrefs Brand Radar (or any source of AI citation counts) connected, ai_visibility · citations puts the pages ChatGPT, Perplexity, Gemini and Google AI cite next to the ones they skip. OnPage.dev scans every page and shows which checks the cited pages pass and the others fail, the averages side by side (AI readiness, words, facts per 100 words, outside sources, entity graph issues), and what each skipped page should change. It is honest about the limits: a correlation over a few pages, not proof of cause.
What prioritize_fixes looks for: blockers on pages with traffic, low click-through for the position, keywords in striking distance (position 4 to 15), top queries missing from the title or H1, and pages with strong backlinks that score badly.
With GA4 conversions (key events) or revenue per page, it ranks by money instead of clicks: a pricing page that converts comes before a busy blog post that does not.
A rank tracker that knows what matters
No keyword lists to maintain. Each run (daily or weekly, from a Claude routine or a dashboard refresh), the assistant passes your Search Console rows and, if connected, your Ahrefs Rank Tracker or Semrush Position Tracking rows to track_rankings. OnPage.dev merges Search Console's average with the tracker's exact position and picks what to follow by itself:
- Work you did: the queries of every page in your fix log, with the day it changed
- Money keywords: pages with GA4 conversions or revenue, commercial or transactional intent, your own list
- Near wins: positions 4 to 20 with real demand
- Top 3 to defend, plus everything you already track in Ahrefs or Semrush
You: How did our rankings do this week?
Assistant: (Search Console) + (Ahrefs Rank Tracker)
track_rankings run 6, 214 keywords followed
Needs attention: "buy boots" off page 1 (8 → 15) [money]
"buy boots" now ranks with /blog/boots instead of /boots
→ diagnose_cannibalization
Wins: "trail shoes" onto page 1 (13 → 5) [money]
Work done: /trail (title, 1 Oct): 9 queries, average +6 places
Every run lists new top 3 and page 1 entries, keywords lost, big movers, a different page ranking (a cannibalization signal), vanished queries and AI Overviews appearing, with the next tool for each, and can post to Slack or Discord. It keeps the first run and the last 90.
Which of your pages wins the keyword
Two pages of yours chase the same keyword and the wrong one ranks. check_keyword · cannibalization reads the page that should rank and the pages that outrank it, live, and puts them side by side on every on-page signal, then says what to change:
You: Our blog post outranks the product page for "trail running shoes". Why?
Assistant: (Search Console) rows blog post at #4, product page at #14
diagnose_cannibalization
| Signal | Product page | Blog post |
| Title | partial | exact @0 |
| H1 | – | exact |
| H2 / H3 | 0 / 1 | 3 / 2 |
| Body mentions | 2 (0.3%) | 11 (1.4%) |
| Bold / italic | 0 / 0 | 4 / 1 |
| Alt / file names| 0 / 0 | 3 / 2 |
1. Strengthen the product page: exact keyword at the start of the title, in the H1 and an H2.
2. Tone down the blog post: keyword out of its title, H1 around its own angle, stop bolding it.
3. Link from the blog post to the product page with "trail running shoes" as the anchor.
It counts the keyword in the title (and how early), URL, meta description, H1, H2 and H3, the body text and the first 100 words, bold and italics, image alt texts and file names, checks canonical and noindex, and maps which of the pages link to each other with the keyword as anchor. When two pages say nearly the same, it proposes a merge with a 301 instead. content · topics finds the competing pages across a whole site first.
Disavow without the risk, and manual actions cleaned up
A wrong disavow file can cost more rankings than the spam ever did. backlinks · disavow takes your backlink data (Ahrefs is_spam, Semrush toxic markers or a CSV) and only puts clear spam in disavow.txt: domains marked as spam, or spam anchors on sites with no authority and no traffic. Anything with authority goes on a list for a person to decide, and without a manual action or real negative SEO it tells you not to upload the file at all.
Got a manual action? explain_google · manual_action covers every type in Google's documentation (unnatural links, thin content, cloaking, structured data, site reputation abuse and more): what Google means, the cleanup steps in order, the tools that check them, what to keep as proof, and a reconsideration request draft with a TODO for everything you actually did.
Bring your own crawl
site_audit · start reads up to 25 pages. Already crawled the site with Screaming Frog, Ahrefs Site Audit, Semrush or Sitebulb? site_audit · import takes up to 500 page rows and the internal links, and returns an audit_id that every audit tool understands: site_audit · continue, internal_links · equity, internal_links · suggest, edge_worker · rules, technical · parameters and share · export.
You: Here is our Screaming Frog export. What should we fix first?
Assistant: site_audit · import 480 pages, 6,912 internal links
site_audit · continue meta description missing on 118 pages, 9 broken pages
internal_links · equity /pricing is 5 clicks deep with 2 links in
edge_worker · rules one rule fixes the descriptions on all /blog pages
Only the columns you send are checked, so a crawl without a description column never reports 480 missing descriptions. The rows live for one hour, long enough to work through the audit. After that only a compact snapshot (address, status, score, title, canonical, noindex and issue names per page) stays for 90 days, so site_audit · compare can show what changed next time.
What changed since the last audit
Every finished site audit, crawled or imported, leaves a small snapshot for 90 days. After the next release, run the audit again and site_audit · compare shows what moved:
You: We shipped the new templates. Did anything break?
Assistant: site_audit · import 480 pages from this morning's crawl
site_audit · compare vs the crawl of 3 October
Regressions: /pricing now returns 404, /blog/guide turned noindex,
12 product pages changed their canonical.
Fixed: meta descriptions on 118 pages, 9 broken pages work again.
Or let it run by itself: watch · site_audit audits the site every day, week or month (up to 25 pages), compares each audit with the last one, and posts the changes to Slack, Discord or any https webhook when pages broke, turned noindex, changed canonical or got new issues. action: "status" shows the last result.
Links to win, and AI answers to enter
Two more jobs where your SEO tool has the data and OnPage.dev makes it a plan:
track_rankings: Search Console plus Ahrefs or Semrush rank tracking, keywords picked from your work, money pages and near wins, with what moved each run.backlinks · disavow: disavow.txt with only clear spam, the rest up for review, and a warning when you should not upload it.explain_google · manual_action: every Google manual action type with cleanup steps, checking tools, proof to keep and a reconsideration request draft.check_links · outbound: dead links to other sites, cited pages that now redirect to a home page and moved pages, with an archived copy as replacement.check_keyword · cannibalization: the page that should rank next to your pages that outrank it, signal by signal (title, URL, meta, H1 to H3, mentions, first 100 words, bold, italics, alt texts, file names, internal keyword anchors), with the plan to fix it.plan · split_test: 20+ pages of one template split into two groups with equal traffic, a Worker that changes the title or description on one group only (same for visitors and Googlebot), measured withmeasure · impact.site_audit · report: a finished site audit as a clean report page with your agency name, to send as a link or save as a PDF.text_quality · spelling: real spelling and grammar errors in a page's main text (LanguageTool, about 30 languages), brand names and style advice left out.sitemap · generate: sitemap.xml from an audit with only indexable, self-canonical pages and the reason for every page left out.- Sections and competitors:
site_audit · continueshows issues per section;site_audit · comparewithcompetitor_audit_idcompares two sites side by side. find_opportunities · content_decay: monthly clicks per page (Search Console, Ahrefs pages-history or GA4) become a refresh queue of pages that peaked and kept falling, seasonal dips left out.get_fix_pack · alt_texts: lists images without alt text; your assistant looks at them and writes the texts; OnPage.dev checks them and ships them as HTML, a Cloudflare Worker, WordPress or code.- Applied internal links:
internal_links · suggestwithplatformreturns a Cloudflare Worker that links the first mention on each page (never inside an existing link), a WordPress plan or code edits.
You: Why does ChatGPT recommend Rivalo and not us?
Assistant: (Ahrefs) Brand Radar 120 answers for us and 2 competitors
ai_visibility · gaps ChatGPT names Rivalo in 41%, us in 12%
14 questions where only they appear
cited most: zapier.com list, g2.com, reddit
backlinks · opportunities zapier.com links to both rivals, not to us
And the links you lost
Other sites still link to pages you deleted or moved, and visitors still land on them. Every one of those links and visits is value you no longer get. With Ahrefs, Semrush, GA4 (page views of your 404 page) or a CSV export, backlinks · broken turns them back into ranking power and traffic:
You: Which of our backlinks point at dead pages? Fix them.
Assistant: (Ahrefs) broken backlinks 112 links to 19 dead pages
backlinks · broken ranked by linking sites and domain rating
top 20 checked live: 14 dead, 3 already fixed, 2 block bots
1. /blog/running-shoes-guide 38 sites, best DR 81 → /guides/running-shoes
2. /old-promo 12 sites, best DR 64 → /socks/trail (anchor text match)
3. /moved reached via 3 redirects → one 301 straight to /new-home
14 rules ready for _redirects. /old-promo is still in your sitemap.
The dead pages are grouped and ranked by link value (each linking site counts once, weighted by domain rating, nofollow at 30%). The top 20 are checked live, so pages that already work again or only block bots are not redirected by mistake. Each dead page gets the best live page from your sitemap or a site audit, matched on the URL, the page title and the anchor texts other sites used. No close match? It says so instead of dumping the link on the home page. GA4 404 hits add real visits to the ranking, and when the visitors came from one of your own pages, that link is listed as an edit instead of hiding it behind a redirect. The rules come out for _redirects, nginx, Apache, Next.js or CSV, with the strongest linking sites worth an email. Then check_links · urls confirms each one is a single 301 and fix_log · log keeps watching them.
Lost backlinks work too. Pass Ahrefs backlinks with their history and lost_reason: a page that died or a redirect that broke on your side gets a 301, and a link the other site took out goes on an outreach list, strongest first, with what to ask ("put it back", "link the new version of your page") and the sentence the link used to sit in.
No Search Console connected? find_opportunities · striking_distance and content · topics also take Ahrefs or Semrush organic keyword rows (keyword, url, position, volume, traffic). And content · topics with competitor_rows gives a keyword gap: topics where competitors rank in the top 10 and you do not, by search volume, with the page to strengthen or the page to write.
Writes in your site's own voice
Most AI writing tools sound like AI writing tools. OnPage.dev does not write at all: your own agent writes, and OnPage.dev gives it the eyes, the plan and the checks, so every post sounds like the rest of your site and is built on what already worked there.
You: Plan next month's posts for example.com and draft the first one.
Assistant: content · voice 5 articles read: sentences ~14 words, paragraphs ~52,
"you" voice, 30% question headings
content · calendar 8 items over 4 weeks
Mon Refresh /guides/compost "compost bin" (pos 11, +140 clicks)
Thu New "raised bed soil" as a list (lists lifted clicks +18% here)
content · draft Voice match 91/100
1 figure without a source: "cuts household waste by 30%"
SEO scan 88/100
Here is the draft. Approve it and I publish it as a draft in WordPress,
log it, and measure it in 4 weeks.
- Learns the voice. Sentence, paragraph and section length, question headings, lists, "you" or "we", title length and sample passages, from 2 to 5 of your best articles. Every content brief then carries it.
- Plans the calendar. Refreshes for queries almost on page 1 and new pages for topics you miss (from
content · topicsorai_visibility · gaps), alternated and dated per week. Keywords already drafted are skipped. - Checks every draft before you see it. Voice match with concrete tips, figures without a source link, and a full SEO scan of the HTML.
- Learns what works on your site. Once a draft is published and logged,
measure · impactresults are tied to its format, length, title and FAQ. Three measurements make a rule, Google updates are left out, and every fifth item tests a format not tried yet. - Keeps the loop going. The calendar is saved in the site memory, so
site_memoryshows what is due and overdue, and which published posts are ready to measure. Add a Slack or Discord webhook and it reminds you every Monday. - Never publishes on its own. The
content_autopilotprompt runs the whole loop and waits for your approval before anything goes live.
Matches search intent
The fastest way to waste a month: write a product page for a keyword where Google only ranks guides. content · intent checks what actually ranks before you write.
Your agent pulls the top results for the keyword from Ahrefs or Semrush, and OnPage.dev scans every one of them and works out the page type Google rewards: guide, listicle or comparison, product page, category page, service page, tool, local business page, news article, recipe, help page, pricing page, forum or video.
You: Does our page fit "best running shoes for flat feet"?
Assistant: (Ahrefs) serp-overview top 7 results
content · intent 8 pages read
Your page fits: Google ranks listicles here (runrepeat, rtings),
and yours is one too.
4 of 7 results are forums (Reddit, JustAnswer): searchers want
real experience. Add first-hand testing, photos and named experts.
43% of the top results put the year in the title.
- Page type match: match, close (same intent, other format) or mismatch, with the evidence per result.
- Format that ranks: median length and subheadings of the top results, how many have a FAQ, a year or a number in the title.
- Forums and videos: recognised from the address, without scraping them, and turned into concrete advice about experience and trust.
- The whole site: pass Search Console rows and it finds queries where two of your pages compete with each other, and pages that attract mixed intents.
Readability that fits the page
A product page should not read like documentation. text_quality · readability measures reading ease (Flesch, or Flesch-Douma for Dutch) against the right target: the median of the pages that rank for the keyword, or the common norm for the page type when there are no competitors. It lists the hardest sentences and paragraphs to rewrite first, ignoring menus, link lists and tables.
Assistant: text_quality · readability Reading ease 63, ranking pages median 52 (52, 63, 49)
Hardest sentence (22 words, score 8):
"Runner's World Running Reviews Editor, Amanda Furrer, who has
relatively low arches, researched shoe testing data from..."
Ships to WordPress
From chat to live, with you approving every change. get_fix_pack with platform: "wordpress" turns the fixes into a WordPress plan your agent can apply through your own WordPress connection (for example a WordPress MCP server):
You: Fix the SEO on /trail-shoes and put it live.
Assistant: scan_page 46: title too short, no meta description, 1 image without alt
get_fix_pack platform: wordpress, 3 changes
SEO title -> Yoast _yoast_wpseo_title
Description -> Yoast _yoast_wpseo_metadesc
Alt text -> media library
Shall I apply these 3 changes? You: yes
(WordPress) update post + media
scan_page · rescan 94, all 3 fixed
- The exact field for each value in Yoast SEO and Rank Math, alt texts for the media library, and structured data as a Custom HTML block.
- Your agent shows every change as old and new value and waits for your approval before anything goes live.
- OnPage.dev never logs in to your site and stores no passwords: your own WordPress connection makes every change.
- Without a WordPress connection, the same plan tells you exactly what to paste where.
One-click approval with the WordPress plugin
Install the OnPage.dev WordPress plugin (source), click Connect in Tools, OnPage.dev and give your agent the site key. Then get_fix_pack with mode: "wordpress" sends the fixes straight to your wp-admin:
New in plugin 1.1: your agent can also send a whole post. A new post or page arrives as a draft you publish yourself. An update to an existing post shows the removed and added paragraphs side by side and goes live when you approve it. Undo restores the old text unless someone edited the post since.
You: Fix the SEO on /trail-shoes and send it to WordPress. Site key wpk_...
Assistant: get_fix_pack · wordpress 3 changes waiting for approval
wp-admin: Tools, OnPage.dev
title Trail shoes -> Trail running shoes for rocky ground | Acme
meta_description (empty) -> Grip, cushioning and sizes for ...
jsonld (none) -> Product
[Approve selected] [Reject all]
Assistant: get_fix_pack · wordpress change_id: applied, old values kept
scan_page · rescan 94, all 3 fixed
- Old and new value per field, approve all or some, undo any change set later.
- Written through Yoast SEO or Rank Math when active, otherwise the plugin outputs the title, description, canonical and social tags itself. JSON-LD and media library alt texts work everywhere.
- Pull only: the plugin fetches proposed changes when you open its page. OnPage.dev never logs in to your site, and nothing changes without an administrator's approval.
Ships to your CMS, not just WordPress
get_fix_pack writes the plan for the platform the site runs on, with the exact MCP tool or API call and field for every fix, applied through the user's own connection after approval:
Ships as a pull request
Not on WordPress? get_fix_pack with platform: "git" writes the fixes as code for the site's framework and opens a pull request through the git tools your agent already has. It detects Next.js (App and Pages Router), Nuxt, SvelteKit, Astro, Angular, React, Vue and plain HTML from the page itself.
You: Fix the SEO on /pricing and open a PR.
Assistant: scan_page 58: title too short, no canonical, no structured data
get_fix_pack platform: git, detected Next.js (App Router)
export const metadata = { title, alternates.canonical }
JSON-LD in the page component
git switch -c seo/onpage-pricing
(edit app/pricing/page.tsx, build)
scan_page · html 91, nothing else changed
gh pr create #42 "SEO: title, canonical, jsonLd for /pricing"
metadatafor Next.js,useSeoMetaanduseHeadfor Nuxt and Vue,<svelte:head>for SvelteKit, theTitleandMetaservices for Angular, head tags for Astro and plain HTML.- Always a new branch, checked with
scan_page · htmlbefore the PR, never merged by the agent. - For client-side React, Vue and Angular apps it also says how to prerender, because AI crawlers do not run JavaScript.
Share cards, with a free API
snippet · social shows how a page looks when it is shared on Facebook, LinkedIn, X, WhatsApp and Slack: every Open Graph and Twitter tag, and the share image itself. It loads the image and reads its real size from the file header, so it catches the cases other checkers miss: an og:image that returns 404, a 400 x 400 logo where a 1200 x 630 image belongs, an SVG that no platform shows, or a 6 MB file.
The same check is a free public REST API, no key needed:
curl "https://onpage.dev/api/v1/og?url=https://example.com"
{ "url": "https://github.com/", "score": 100,
"image": { "ok": true, "type": "image/png", "width": 1200, "height": 630, "bytes": 618778 },
"issues": [] }
10 requests a minute per IP address, CORS open, each result cached for 5 minutes. Try it in the browser: Open Graph checker.
Keeps watching after you ship
Most SEO problems are not there on launch day. They arrive later: a deploy adds noindex, a CMS update changes the canonical, a new robots.txt blocks GPTBot. watch · page rechecks a page about once a day and records every change for search and AI. Alerts go to a private RSS feed and, if you give one, a Slack or Discord webhook. No account, no key.
You: Keep an eye on our pricing page and tell me if anything breaks.
Assistant: watch · page baseline 94/100, 8 of 8 AI crawlers allowed
...three days later, in your Slack:
OnPage.dev: 2 change(s) on https://example.com/pricing
- A noindex tag was added: the page will drop out of Google
- The score went from 94 to 71
And three more jobs agents get asked to do every week:
content · brief: give a search term and the pages that rank, get a writing brief: the subtopics most of them cover, the questions they answer, figures to back up with your own data, a target length, schema and internal links. Bot walls and navigation headings are filtered out.plan · redirects: give the old and new URLs (or the sites, to read their sitemaps) and get a 301 map with a confidence score per match, written as_redirects, nginx, Apache, Next.js or CSV. Then verify after launch withcheck_links · urls.site_audit · import: pass the page rows of a crawl you already have (Screaming Frog, Ahrefs Site Audit, Semrush, Sitebulb), up to 500 pages plus the internal links, and every audit tool works on the whole site. Only the columns you send are checked.site_audit · compare: a finished site audit against the previous one of the same site (snapshots kept 90 days): pages that broke, turned noindex or changed canonical or title, issues new or fixed per code, pages added or gone, average score before and after.watch · site_audit: audits a whole site every 1, 7 or 30 days, compares each audit with the previous one and posts regressions to a webhook (Slack, Discord or any https endpoint). Status and stop with the schedule_id.backlinks · opportunities: competitor backlink gap (sites linking to 2+ competitors, not to you) ranked and typed with how to approach each, plus unlinked brand mentions read live with the sentence to quote.ai_visibility · gaps: real AI answers (Brand Radar or similar) turned into share per engine, questions where only competitors are named, the AI's searches and the sites it cites.backlinks · broken: pass the broken backlinks from Ahrefs, Semrush or a CSV export. The dead pages are ranked by the sites that still link to them, the top 20 are checked live, each gets the best live page to 301 to (from the sitemap, a site audit and the anchor texts), and the rules come out ready to ship. Dead pages still in the sitemap, redirect chains and temporary redirects are flagged too.share · link: after fixing a site, publish a before and after page (for example 46 → 94, with the issues fixed) to send to your client. Not indexed, expires after 90 days.
Remembers every site
Agents forget everything between chats. OnPage.dev does not. site_memory keeps one memory per site, behind a secret memory_key: open it in a chat and every OnPage.dev tool call is recorded (what was found, shipped and measured), and guards, scheduled audits, competitor watches and fix logs started in that chat add their results too. The next chat, in Claude, ChatGPT or Cursor, starts where the last one stopped.
You: Pick up where we left off on example.com.
Assistant: site_memory · open 41 steps since July: pricing 46 → 91, 12 fixes shipped,
+38% clicks vs control, guard caught a noindex on 3 Oct
Next: 1. Measure the 4 fixes shipped 5 weeks ago
2. /blog/guide scored 58, no fix written yet
3. No rank tracker yet
It learns. Every measured result is tied to the fixes shipped before it, so with a memory open prioritize_fixes, get_fix_pack and plan · sprint show which fix types moved traffic on this site ("title rewrites +38% here"), and site_memory · next suggests rolling them out to more pages. Decisions saved as notes come with every fix pack, brief and draft.
It connects the dots. A ranking drop, traffic drop, decay, vitals regression or alert comes with the fixes, deploys and decisions of the last three weeks, so the agent checks those first.
It learns across sites, if you opt in. Open the memory with share_learning: true and each measured result adds one anonymous row per fix type to a shared pool: the fix type, the change rounded to a whole percent and the month, never the site, addresses, numbers or text. Once a fix type has 20 measurements from 10 sites, prioritize_fixes, get_fix_pack and plan · sprint show the median change across sites. Turning it off or deleting the memory removes everything that memory added.
The memory stores summaries and a few numbers per step, not your raw data rows or page content. The key is stored only as a hash, and an unused memory is deleted after 400 days.
Acts, guards and measures like an SEO team
The work around the fixes, done by your agent with the tools you already connected.
You: We shipped the new templates. Make sure nothing broke and plan next month.
Assistant: request_recrawl 12 of 12 pages indexable, IndexNow sent
watch · guard guarding robots.txt, sitemap and 5 key pages
plan · sprint 19 tasks over 4 weeks, tickets ready for Linear
...two days later, in your Slack:
OnPage.dev guard: /pricing has noindex. Fix: remove the noindex meta tag.
Built for agents
What makes OnPage.dev different from every other SEO MCP server:
- Everything in one chat, dashboards included. Data, plan, the fix your agent deploys and the proof all happen in the chat.
dashboard · feedchecks a site's key pages every day,dashboard · datareturns the history as clean tables that join with GA4 and Search Console, anddashboard · builddesigns a live dashboard for Claude with a realtime GA4 block (active users per minute, pages, events) next to SEO health, fixes, speed and AI visibility. - Takes action, not notes. Your agent ships the fix itself: a pull request in your framework, a change through your own WordPress connection, or a Cloudflare Worker from
get_fix_packwithplatform: "cloudflare"that corrects the tags as the page is served, on any CMS on your own Cloudflare account. One step to roll back. - Closes the loop for real.
fix_log · logrecords every fix that shipped. OnPage.dev checks about every 6 hours that each one is still live, sends the restore code to your webhook when a deploy undoes it, savesmeasure · impactresults per fix with a rollback plan when one did worse, andfix_log · reportwrites the client report. - Proves the result.
measure · impactcompares before and after numbers against a control group of unchanged pages. More. - From finding to done.
share · exportsends the plan to Google Sheets (no connector needed), Slack, Notion or a task board. More. - Sees every reader. First screen on phone, tablet and desktop in a real browser, and the page read aloud like a screen reader. More.
- Handles JavaScript apps. Pure client-side React, Angular and Vue apps get a full audit of the rendered page, the way Google sees it, plus how little AI crawlers get without JavaScript.
- Ships anywhere. WordPress through your own connection, every other stack as a Git pull request with framework code. More.
- Real users and real bots.
web_vitals · nowreads Core Web Vitals from real Chrome users;technical · logsshows what Googlebot and AI crawlers actually crawl. - Multilingual and trustworthy.
technical · hreflangchecks every language version and writes the corrected tag set;entity · trustchecks the E-E-A-T signals across the whole site.
Install in one minute
One click for Cursor and VS Code. For Claude, ChatGPT and other clients, see below.
The server is hosted. There is nothing to install or run, and no key.
Fewer tools in context: use https://onpage.dev/mcp?tools=core for the 8 main tools (site memory, scan, AI visibility, fix pack, layout, site audit, prioritise, share). The full server has 37 tools, under the roughly 40 that Cursor shows across all servers..
Server URL: https://onpage.dev/mcp
claude mcp add --transport http onpage https://onpage.dev/mcp
Or install the plugin for the server plus ready commands and skills.
Settings → Connectors → Add custom connector → paste https://onpage.dev/mcp → save. Turn the connector on in a chat.
Add a custom connector with https://onpage.dev/mcp (Settings → Connectors, with developer mode on). Which plans can use custom connectors is up to OpenAI.
~/.cursor/mcp.json or .cursor/mcp.json in your project:
{
"mcpServers": {
"onpage": { "url": "https://onpage.dev/mcp" }
}
}
.vscode/mcp.json:
{
"servers": {
"onpage": { "type": "http", "url": "https://onpage.dev/mcp" }
}
}
~/.codeium/windsurf/mcp_config.json (Windsurf uses serverUrl), then Refresh under Manage MCPs:
{
"mcpServers": {
"onpage": { "serverUrl": "https://onpage.dev/mcp" }
}
}
~/.gemini/settings.json, then check with /mcp:
{
"mcpServers": {
"onpage": { "httpUrl": "https://onpage.dev/mcp" }
}
}
codex mcp add onpage --url https://onpage.dev/mcp
Or [mcp_servers.onpage] with url = "https://onpage.dev/mcp" in ~/.codex/config.toml.
cline_mcp_settings.json:
{
"mcpServers": {
"onpage": { "type": "streamableHttp", "url": "https://onpage.dev/mcp" }
}
}
settings.json (a small bridge with Node.js):
{
"context_servers": {
"onpage": { "command": "npx", "args": ["-y", "mcp-remote", "https://onpage.dev/mcp"] }
}
}
MCP Client Tool node on an AI Agent: Server URL https://onpage.dev/mcp, transport HTTP Streamable, no authentication. Workflows can then scan, audit and report on a schedule.
Streamable HTTP transport at https://onpage.dev/mcp, no authentication. Supports MCP 2026-07-28 (stateless, server/discover) and the earlier revisions with a session down to 2024-11-05. Without sessions, pass memory_key on tool calls to record them in the site memory.
What it can do
37 tools that cover 116 jobs. Related jobs share one tool and are picked with mode, for example site_audit with start, continue, import, compare and report; leave mode out for the first one. Old tool names from before 2.0 keep working. Every result links to the full visual report on onpage.dev. Issues carry stable codes; the full list is in the onpage://rules resource.