Monitor any RSS feed, turn new posts into platform-tailored social drafts in a Notion content queue, learn from your edits, and schedule approved posts

On first run the agent asks the user:

• (1) which RSS feeds to watch (the user pastes any number of feed URLs — newsletters, blogs, podcasts, news sites),
• (2) the destination for the content queue (Notion is the default because it is the most common choice for solo creators, and the agent sets up a database with fields: title, source URL, source published date, AI summary, draft per platform, status — draft / approved / scheduled / posted, scheduled time, posted link, edit diff). The user can pick a different queue (Google Sheets, Airtable) at any time,
• (3) which social platforms to support — any combination of Twitter / X, LinkedIn, Facebook, Instagram, Threads, Mastodon, Bluesky. For each platform the user enables, the agent checks whether the corresponding account is connected (via Zapier or Composio). If a platform is not yet connected, the agent still generates a platform-tailored draft (right length, right tone, right format conventions — e.g. a thread for Twitter, a longer reflective post for LinkedIn, an image-led caption for Instagram) and stores it in the queue with the status "draft (platform not connected)" so the user can copy-paste manually.
• (4) The voice baseline — the user can either let the agent default to a clean informative tone, paste 3-5 example posts they like, or write a short voice brief in plain language.
• (5) Daily or weekly delivery of new drafts to the Notion queue (user picks) and the post-schedule cadence (e.g. 3 posts per day spread across platforms, or only on weekdays, or only LinkedIn on Tuesdays — fully user-configurable).

These preferences are stored.

On every fire the agent pulls the latest entries from each enabled RSS feed, runs a single AI pass to produce a tight summary, then generates one draft per enabled platform with the platform's length, tone, and format conventions applied. Each draft is created as a row in the Notion content queue with the source link and AI summary attached. The user reviews in Notion — they can edit any draft, change the status to approved, and set a scheduled time. The agent then posts the approved draft to the platform via Zapier or Composio at the scheduled time. A feedback loop is built in: after every user edit, the agent stores the diff between the AI draft and the user's final version, and uses that diff to refine its future drafts. A weekly recap shows the user how many drafts were approved with zero edits vs edited, the most common edit patterns (e.g. "user always shortens Twitter drafts by 20%"), and the implied prompt refinements the agent has applied.

The user can at any time ask the agent for an on-demand view ("show me this week's drafts", "what did I edit most", "regenerate the LinkedIn version of this one with my voice"), change the feeds, add or remove a platform, switch the queue, or update the voice baseline.