Git publishing
How ConveyLoop publishes SEO pages to your Git repository
ConveyLoop commits finished pages to one folder in your repository. Your host deploys them like any other commit, and ConveyLoop checks that each page is live.
What you connect
A repository on GitHub.com (through the ConveyLoop GitHub App), GitLab.com or Bitbucket Cloud. You choose the repository and the branch the pages go to. Self-hosted editions (GitHub Enterprise Server, self-managed GitLab, Bitbucket Data Center) aren’t supported yet. A project connects one repository.
What ConveyLoop writes
Everything goes in one folder, the Target subdirectory: /conveyloop/seo unless you change it. ConveyLoop only creates, changes or deletes files inside it. It never reads or changes the rest of your repository.
conveyloop/seo/
├── content/
│ ├── _index.html the section page, at /guides
│ ├── example-page.html a page, at /guides/example-page
│ └── conveyloop.css the stylesheet the pages link
├── sitemap.xml these pages' addresses
├── conveyloop-manifest.json ConveyLoop's record of what it wrote
└── README.md- One file per page, named after the page: lowercase letters, digits and hyphens.
_indexis the section’s own page, listing the others. It appears once there are pages to list, when the pages are served under a path such as/guides.- The manifest is how ConveyLoop removes a page you unpublish. It never deletes a file it didn’t write.
- Don’t edit these files by hand. ConveyLoop rewrites them whenever it publishes, so change a page in ConveyLoop instead.
Where your site serves the pages
A page’s address is separate from its file. ConveyLoop picks the path the pages are served under: /guides, or /learn or /resources if your site already uses /guides. You can change it in the Git window. Each page’s canonical link, its sitemap entry and the links between pages all use that address.
Your site’s routing connects the two, so it needs a one-time change to answer /guides/example-page with conveyloop/seo/content/example-page.html. On most sites that is one route or one rewrite rule. A host won’t do it on its own, because the files sit in content/ inside the folder rather than at the path.
How changes land
- Commit directly (the default) is fully automated: approved pages commit straight to your branch.
- Open a pull request (a merge request on GitLab) puts each publish up for review before it ships.
If the branch is protected, it refuses direct commits until you allow them: on GitHub, add the ConveyLoop app to the branch’s bypass list; on GitLab and Bitbucket, the account you connected with must be allowed to push to it. Or use pull requests.
Page formats
Standalone HTML (the default)
Complete pages, ready to serve as they are, with ConveyLoop Metrics built into every page, so there is nothing to install. Styling comes as one conveyloop.css beside the pages (the default), inside each page, or not at all. Every rule sits in a CSS cascade layer, so your own CSS always wins over it.
MDX or Markdown
Content files your site renders in its own layout, for frameworks like Astro and Hugo. Each opens with front matter: title, meta_title, meta_description, canonical_url, published_at, slug, and category when the page has a topic. Content files can’t carry the metrics script, so add the ConveyLoop Metrics snippet to your layout or connect Search Console: ConveyLoop needs one of them to measure the pages.
Deploys
Your host builds and deploys when the commit lands, as it does for any push: Vercel, Netlify, Cloudflare Pages, Render and the rest. If yours doesn’t deploy on push, add a deploy hook in the Git window: a Netlify build hook, a Cloudflare Pages deploy hook, or any HTTPS address that starts a rebuild on POST. ConveyLoop calls it after a direct commit, or when its pull request is merged.
ConveyLoop checks every publish goes live
About three minutes after a publish lands, ConveyLoop fetches the newest page at its address. A page that isn’t there yet is checked again ten minutes later, then thirty minutes after that, because a build can be slow.
If the page is still missing, or your site turns ConveyLoop away, the Git window says so and you get an email. Both carry a prompt to paste into your AI coding agent in your site’s repository. It makes the one-time change that serves the pages, so nobody has to write a route by hand.
ConveyLoop’s requests identify themselves as ConveyLoopBot and ConveyLoop-Crawler. If your site has bot protection or a firewall, let those through for your public pages, /robots.txt and /sitemap.xml (or the sitemap your robots.txt names). The crawl starts at your homepage, so allowing only the pages’ path isn’t enough. Keep the protection for everything else, such as login and account pages: a User-Agent is easy to fake.
Getting the pages into search
List the pages in your site’s sitemap: the one at /sitemap.xml, or one your robots.txt names. Search engines find them there, and ConveyLoop’s own crawl reads the same sitemap. The folder’s sitemap.xml already lists them; if your site serves that file, a Sitemap: line in robots.txt pointing to it is enough.
Disconnecting
Disconnect at any time. ConveyLoop opens a pull request (a merge request on GitLab) to remove its files, and never deletes them directly.
Setup guides
Step-by-step setup for a specific framework or host. We publish one only after running it end to end.
Using something else? Connect anyway. If your site doesn’t serve the pages, the check finds it and gives you the prompt that fixes it.
Publish SEO pages from your own repository
ConveyLoop writes, commits and improves the pages. Your site keeps its own code and its own deploys.