When your site is built from a repository (Astro, Next.js, Hugo, Jekyll, Docusaurus, Gatsby and similar), Rank Prompt connects to it as your site. Website autopilot changes the real files of its pages, and that needs no blog. If the repository has a blog, Rank Prompt also reads how its posts are written (the posts folder, the file name, the frontmatter keys, images, languages and the address a post goes live at) and writes each article the same way.
Everything arrives on a new branch as a pull request you review. The change goes live when you merge it. The one exception is full autonomy in Website autopilot, where Rank Prompt merges its own pull request once the repository's checks pass, and leaves it for you when a check fails.
The repository connected on the GitHub card in Integrations > Publish content (the + on the card; the gear on the repository row changes the repository, the folder the agent may touch and the account)
Starter or a higher plan on the brand owner, and you must own the brand to set it up
Open Integrations > Publish content. The GitHub card names the connected repository and says Repository not read yet. Without a repository, click + on the card to connect one.
Click Read the repository. Rank Prompt reads it and shows Site connected with the framework and the site's address. With a blog it also shows Blog ready with the posts folder, and How each article lands in this repo:
Saved as: the file a post becomes, like src/content/blog/{locale}/{slug}.mdx
Sent as: a pull request into your branch, or a direct commit
Published at: the address of a merged post, like {site}/blog/{slug}
the frontmatter a post carries, how images are handled and the languages
What we read lists the files it read and how sure it is.
Click Edit to correct anything. Repository settings open with the site (Site URL, Branch), where posts go (File path, Live URL) and how they arrive (Pull request or Direct commit); format, images, languages and frontmatter sit under one more click. Read again later replaces what was read but keeps your corrections, the branch and how posts arrive.
A site address is needed to match your pages and give a published post its link. If none was found, the card says Site address missing: add it with Edit.
Editors of the brand see the card and what was read, without the buttons.
A repository without a blog still connects as your site, and Website autopilot works on it. Only articles need a blog: the card says No blog, and the repository is not offered as a place to publish them. For Astro, Next.js, Hugo and Jekyll, Add a blog opens a pull request that adds one following the repository's own layout, and articles publish here once it merges. For other frameworks, Edit lets you say where posts go.
Open an article, click More actions (the three dots) and Publish, and pick the repository.
Publish opens a pull request. The success window says Pull request opened with a Review PR button, and the publish window shows Waiting for merge until it merges.
Save as draft opens a draft pull request, to preview the post. Publishing the article later marks it ready.
Schedule opens nothing until the time comes.
Publishing again while the pull request is open adds a commit to it. Publishing a live post again opens a new pull request that updates the same file.
With Direct commit as the delivery, a live publish is committed to the branch and goes live with your next deploy, with nothing to merge.
A pull request closed without merging leaves the article as a draft. On the Articles list, an article waiting for its merge shows the Review pull request action, and the Calendar shows it as Waiting for merge.
When Rank Prompt reads your repository it also looks for where your site keeps pages, and a city page from Business Profile > City pages goes there as a file at its planned address, through the same pull request or direct commit as a post:
Astro: the folder of the content collection a catch-all route (src/pages/[...slug].astro) renders, or Markdown pages in src/pages with the layout your existing Markdown pages use.
Next.js: app/{address}/page.mdx when your config takes MDX pages, or the content folder a catch-all route (app/[...slug]/page.tsx) reads.
Docusaurus: src/pages/{address}.md.
Jekyll: {address}/index.md with your page layout and the address as its permalink.
The file carries the page's title and description the way your other pages do. A file already at that path that Rank Prompt did not write is never replaced or renamed around: the publish stops and names it. A repository where none of these is found (Hugo and Gatsby among them) takes the city page as a blog post.
A repository works as the destination of Content > Autopilot: each article waits in its pull request and counts as published once it merges. Links back from your related posts cannot be edited into the repository from here, so they come as tasks with the sentence to paste.
The agent can read the repository for you ("read my repo") and open the pull request that adds a blog ("add a blog to my site"). Both need the brand owner.
Check that the Rank Prompt GitHub App still has access to the repository, then publish again.
Correct Live URL or Site URL with Edit; the next publish uses it.
Email [email protected] and I'll help.
