Ship fast. Stay indexed.
RankForge reads your site the way Googlebot does and checks it where SEO breaks: in your editor, in every pull request, and in production. It fails the pull request that would drop pages from Google, and hands each fix to your AI editor.
Or straight from your terminal, nothing to install
npx rankforge-cli audit your-site.com
Next.js / Nuxt / Astro / SvelteKit / Remix / MDX
RankForge · Critical: 1 critical change: /pricing became noindex (robots meta tag)
Compared preview-412.acme.dev with acme.com. Score 91 → 84.
- critical/pricing became noindex (robots meta tag)
- warning/docs canonical changed to /docs/v2
Fixed: Pages missing a meta description
3 differences that come from the environment, not the code
From code chaos to SEO clarity
Scroll, and watch a messy repo resolve into clean, indexable structure. Every warning turns into a check, and every fix is backed by a diff.
Before: raw repo
- <title> missing
- alt="" on 37 images
- canonical absent
- sitemap: 12 / 142 urls
- h1 × 3 on homepage
- schema.org: none
After: fixes merged
- <title> unique per page
- alt text generated
- canonical valid
- sitemap: 142 / 142 urls
- single h1 enforced
- schema.org: valid
Finding problems was never the hard part.
Crawlers surface hundreds of warnings, weeks after the deploy that caused them. RankForge checks each change as it ships, groups findings by cause, and ranks them by what they cost you in search.
Surface real SEO issues
Detect missing metadata, indexation problems, broken internal links, schema gaps, and performance regressions without drowning in raw crawl noise.
Connect cause and context
One finding per cause, with every affected URL and the files to look at, instead of treating each URL as its own problem.
Prioritize by impact
Rank fixes by search impact, implementation effort, affected URLs, and risk, so the next action is obvious.
Three checkpoints. One engine.
SEO breaks in the same few places: a template change, a config flag, a CDN rule. RankForge checks each moment with the same deterministic engine, so the editor, the pull request and production always agree.
Before you push: your editor
The CLI or the Claude Code plugin audits localhost. Your AI assistant fixes what it finds from the evidence, then compares your version with production.
In the pull request
The GitHub Action audits the preview deployment against production, page by page, comments the result on the pull request, and fails it when pages would leave search.
After the merge: production
RankForge checks the live site on a schedule and alerts you by email, Slack or Discord when something critical changes: a noindex, a failing robots.txt, a page that starts returning errors, a certificate about to expire.
Fix from the evidence
Every finding comes with the affected URLs and the files to look at. Mechanical ones come with a ready patch for your framework.
From crawl finding to a diff you can merge.
RankForge hands your AI editor the evidence: the pattern, every affected URL and the likely files. The editor finds the template in your code and writes the change; mechanical fixes come as a ready patch.
Crawl finding
Missing or duplicated metadata detected across product category pages.
In your code
Your AI editor follows the evidence to the template, CMS field or component behind the pattern.
The diff
A minimal change your team can read, edit and merge. The pull-request check verifies it.
From scattered findings to a pull request that passes.
Related SEO issues are grouped by cause and turned into one focused change, instead of a backlog of tickets. Then the pull-request check proves the fix worked.
Category metadata
Canonical rules
Structured data
Adds metadata fallbacks for category pages with missing descriptions and normalizes canonical generation.
One finding per cause
A repeated issue is reported once, with every affected URL, instead of one ticket per page.
Your pull request
The fix ships the way your team ships everything: a small PR, written by you or your AI editor.
Checked before the merge
RankForge re-audits the preview: the issues must clear, and nothing else may break.
One site watched for free. No credit card.
Technical SEO, without the spreadsheet
54 checks for what actually breaks: indexing, canonicals, sitemaps, structured data, speed and AI-search access. Every finding cites its evidence.
Crawls like Googlebot
Links first, then every URL your sitemap lists. robots.txt is read and honoured the way Google reads it.
noindex, wherever it hides
Meta tags and X-Robots-Tag headers, so a CDN rule or a staging flag can't slip through.
Canonicals and hreflang
Canonicals that point at redirects or errors, hreflang codes Google ignores, missing return links.
Structured data that qualifies
Checks the properties Google requires for rich results, not just that JSON-LD exists.
Core Web Vitals of real users
LCP, INP and CLS from Chrome UX Report data, watched from one check to the next.
Ready for AI search
Which of 21 AI crawlers can reach you, a checked llms.txt, and the signals answer engines quote.
A check on every pull request
Preview against production on each push, with the result commented where your team reviews.
Reports in your name
White-label HTML reports for client sites, and one alert address per client on Studio.
It reads your framework's conventions
RankForge knows where metadata lives and how routes are generated in each stack, so its Next.js, Nuxt and Astro SEO fixes look like code you'd have written yourself.
Fix it before you push
Install the Claude Code plugin, or connect any MCP client. Your assistant audits your dev server, fixes the issues in the repository you already have open, then compares your version with production. RankForge measures; your assistant writes the code.
- Audits localhost. No deploy, no waiting. See issues as you build.
- Fixes in place: your assistant edits the real files and you review the diff.
- Deterministic measurements. Every score comes from the crawl itself, with the URLs to prove it.
How SEO breaks, and where it's caught
Site audits find what is already broken in production, weeks later. RankForge stops it in the pull request, then keeps watching after the merge.
| What happened | Where it's caught |
|---|---|
| A template change sets noindex on /pricing | In the pull requestthe check fails before the merge |
| A refactor drops the canonical from 40 pages | In the pull request |
| hreflang points at a page that now redirects | In your editor |
| A CDN rule sends X-Robots-Tag: noindex on every page | In productionat the next check |
| robots.txt starts answering 500, so Google pauses crawling | In production |
| A page quietly starts returning 404 | In production |
| The HTTPS certificate expires in 9 days | In productionbefore it expires |
| LCP gets worse for real users after a release | In productionfrom Chrome UX Report data |
It reads. You merge.
RankForge only ever reads your rendered pages. Changes reach your code through your own pull requests, checked before they ship.
Read-only by design
RankForge reads your rendered pages, like any crawler. It never writes to your repository.
You open the pull request
Fixes reach your code through your own PR, reviewed like any other change.
Nothing ships unchecked
Patches are proposals. Whoever applies one, the pull-request check verifies the result.
No access to secrets
It reads public pages and your preview, never your source code or your environment secrets.
A history you can trace
Every check is kept with what changed, so a regression leads back to the deploy that caused it.
Open-source engine
The audit engine is MIT-licensed on GitHub: read exactly what it checks, or run it yourself.
Free on your machine. Paid when RankForge keeps watch.
Audits, the editor agent and the CI gate cost nothing. You pay for RankForge checking your production sites every day and telling you when a deploy breaks something.
Free
Everything that runs on your machine, plus one site watched every week.
- Unlimited audits from the CLI and your editor (MCP), up to 500 pages
- CI regression gate: a pull request that breaks SEO fails, preview vs production
- AI-search readiness, llms.txt generator and ready fix templates
- White-label HTML reports with your name and colour
- Monitoring: 1 site, checked weekly, email alerts
Pro
RankForge watches your production sites and tells you when a deploy breaks SEO.
- Monitoring: 5 sites, checked daily, up to 100 pages each
- Alerts by email, Slack, Discord or webhook when something critical changes
- Watches noindex, robots.txt, canonicals, pages that start failing, redirects, the sitemap, the HTTPS certificate and Core Web Vitals
- 12 months of check history and change log
- Everything in Free
Studio
For freelancers and small agencies looking after client sites.
- Monitoring: 25 sites, checked daily, up to 250 pages each
- A separate alert address and webhook per site, yours or your client's
- 24 months of check history
- Everything in Pro
Billing isn't open yet: during the beta, Pro is free on every account. The prices shown are what plans will cost when billing opens, and you'll be told before anything changes.
Not available yet, and not included in any plan until it ships: google search console data (clicks, impressions, index status) next to each issue, scheduled white-label reports emailed to your clients, team seats.
Questions you're probably about to ask
What does RankForge actually do?
It checks your site's technical SEO at the three moments it breaks: in your editor before you push (the CLI or the Claude Code plugin), on every pull request (a GitHub Action that compares the preview with production and comments the result), and in production (scheduled checks with alerts). The measurements are deterministic: no AI guesses, and every finding cites its evidence.
Does RankForge replace SEO experts?
No. It catches the technical breakage that slips through code review: noindex, canonicals, robots.txt, sitemaps, structured data. Strategy, content and links stay with people.
Which frameworks are supported?
Any site that serves HTML can be audited. Ready patches are written for Next.js, Nuxt, Astro, SvelteKit, Remix, Vite + React, MDX and static sites, with the deepest support for Next.js.
Can RankForge change my code or my main branch?
No. RankForge never writes to your repository. It reads your rendered pages; fixes reach your code through your own pull requests, written by you or your AI editor.
Does it need access to my repository?
No. The CLI and the editor agent run on your machine, the GitHub Action runs in your own CI and uses the job's token only to comment on the pull request, and the hosted monitoring only fetches your public pages.
How does the pull-request check work?
Add the GitHub Action with your preview URL and your production URL. On every push it audits both, matches pages by path and comments the changes on the pull request. It fails the check when pages would leave search: a new noindex, pages that start failing, robots.txt blocking Google. Differences that only come from the preview environment, like its noindex header, are listed apart and never fail it.
What does monitoring watch?
noindex, robots.txt, canonicals, pages that start failing or redirecting, the sitemap, the HTTPS certificate and the Core Web Vitals of real users. Each check is compared with the previous one, and you're alerted by email, Slack, Discord or webhook when something critical changes.
What kinds of SEO issues does RankForge find?
54 checks: metadata (titles, descriptions, OpenGraph), indexing (noindex, canonicals, hreflang, soft 404s, robots.txt, sitemaps, HTTPS), page structure, images, structured data against Google's required properties, internal linking, performance and Core Web Vitals, plus 12 AI-search readiness checks.
How is RankForge different from Screaming Frog or Ahrefs?
Crawlers and SEO suites audit what is already live. RankForge runs its checks before a change ships, on localhost and on the preview of every pull request compared with production, then keeps watching production. It doesn't do backlinks, rankings or keyword data, which need a search index RankForge doesn't have.
Does it work with GitLab or Bitbucket?
The CLI works in any CI: npx rankforge-cli diff <production> <preview> exits 1 on a critical regression and can write a Markdown summary. The pull-request comment is built for GitHub today.
Will this improve my Google rankings?
No tool can honestly promise rankings, and RankForge doesn't. It keeps the technical layer working so your pages can be crawled, rendered and indexed correctly, which is the part of SEO your code controls.
Does it support agencies?
Yes, on the Studio plan: 25 monitored sites, an alert address and webhook per client, and white-label HTML reports with your name and colour. Scheduled client reports are on the roadmap.
Does it work with monorepos?
Yes. RankForge audits URLs, not repositories: point it at the app's preview and production URLs wherever its code lives.
Are the patches safe to apply?
Mechanical patches (robots.txt, sitemap, canonical, viewport, lang, JSON-LD, OpenGraph) are written for your framework and come with notes on what to adapt. Apply them on a branch and let the pull-request check confirm the issue cleared and nothing else broke.
Can I try RankForge without an account?
Yes. The free audit at /audit crawls 3 pages of any public URL and shows every measured issue with evidence, no signup. For unlimited audits (including localhost) use the open-source CLI: npx rankforge-cli audit <url>.
Can I use RankForge from my editor?
Yes. Install the Claude Code plugin, or connect the MCP server to Cursor, VS Code or any MCP-capable assistant. Your coding agent audits localhost or production, pulls fix templates, patches the repository it already has open, and compares the result with production. See the guide at /docs/agent.
Make your next deploy SEO-safe
Run a free audit, add the pull-request check in two minutes, and let RankForge watch production. Free for one site, no credit card.