Publishing
One publishing target per site. SEOAgent renders the article, sends it, and records the URL it came back with.
WordPress
Your site URL, your username, and an Application Password (Users → Profile → Application Passwords). No plugin needed. Tags are matched or created, and the slug is looked up first so a retry updates rather than duplicates.
Ghost
Create a custom integration and paste its Admin API key in the form id:secret. Posts carry Ghost's native meta title and description, and the cover becomes the feature image.
Webflow
A site token plus the collection id your posts live in. Because Webflow collections are user-defined, you also tell SEOAgent which field slugs hold the body, the summary and the cover image - "Save and test" checks the body field actually exists and lists the collection's fields if it does not. Items are created live.
Shopify
A custom app with write_content, its admin access token, and your myshopify.com domain. A store can have several blogs, so testing the connection lists them with their ids; pick the one you publish to. SEO title and description are written as the standard global metafields.
Payload
Your Payload base URL, an API key from an auth-enabled collection, the collection posts live in, and the field that holds the body. Payload bodies can be rich text or HTML depending on how the collection was defined, so you choose whether SEOAgent sends HTML or markdown. A rich-text (Lexical) field is not supported.
Webhook, Next.js, Framer and static sites
The webhook target signs every delivery (X-SEOAgent-Signature: t=…,v1=…) so your receiver can verify it, and the JSON feed at /api/public/sites/{siteId}/articles serves published articles with rendered HTML, markdown, meta, FAQs and the cover image URL. Framer has no write API, so a Framer site reads the feed rather than being published to.
Auto-publish
Per site. It switches on each time a target is saved and passes its connection test, and you can turn it off in Settings. When it is on, each scheduled article gets one delayed job that fires at its scheduled time. It is not a recurring scan and it spends nothing.
Export
No target, or a reviewer without an account? Every written article downloads from the editor's Export row as a markdown file (frontmatter with title, description, tags, date and canonical URL, then the body) or as a PDF rendered on our side. Both are built from the saved article, so wait for “Saved” before you download.