Add up to 10 X (Twitter) handles per brand kit. J77 scans them every 6 hours and surfaces recent posts as quick-start ideas on the build page, learning over time which voices produce content that performs for your brand.
When you add a handle to a brand kit, the J77 scanner pulls that account's last 20 original posts (no retweets, no replies) every six hours. New posts show up on your build page next to the Quick Brief, with engagement metrics and a one-click “Use this idea” button that pre-fills your brief.
When you generate content from one of those ideas, the build is tagged with the originating handle. After several builds the closed-loop learning engine starts deriving per-handle effectiveness scores — i.e. it figures out that posts from one source consistently produce content that performs better than another, and biases future generation accordingly.
@levelsio) and click AddNote: X handles are per-brand-kit, not per-account. Each brand can watch a different set of voices appropriate to its audience.
Each source has individual Pause, Resume, and Remove controls. Pausing keeps captured ideas but stops new ones; removing wipes the source from the brand kit (the captured ideas remain in the build page's inbox until used or dismissed).
On New build, the “Ideas from your sources” card shows up to 10 unused ideas — newest first. Each idea displays:
@levelsio)Two actions per idea:
Builds inspired by a monitored source carry an idea_source.kind = ‘monitored_source’ tag plus the source kind/value. The build detail page shows an Inspired by panel with the original post inline.
Once enough builds accumulate (typically 3+ per handle), the learning engine emits source_weight parameters keyed on x_handle:@handle. These show up on the brand kit's Learning page and start influencing generation: prompts begin emphasizing the topical patterns from high-performing sources.
Typo or recently-suspended account. We hit the X API at add-time so this catches it before saving.
The handle posted only retweets/replies in the last 20-post window (we exclude both). Wait for them to post original content, or pick a more active handle.
Each idea is keyed on (target, source, post_id). If a post truly has the same id, it's deduped. If you're seeing near-duplicates (same content, different ids — e.g. an X user reposting themselves), that's an X-side phenomenon we don't collapse.
Admins can verify scan health at /admin/cron-health. The queue is named monitored-sources-scan, runs every 6h. The owner can also trigger an immediate scan via the brand kit's Scan now button.