Every feed and subreddit anyone tracks is read once, on a shared round-robin loop, whatever your configured Sources subscribe to. A busy platform round-trips faster; the schedule never depends on how many brands are watching the same feed.
New items get scored against your keywords and persona, and a comment is drafted in your voice for the ones worth replying to, into your queue.
Use Scan now on Sources to skip the wait for a feed you just added.
The reply box is the same one the Inbox uses. Draft writes a comment from a direction you give it, Polish cleans up one you typed, and the Draft menu lets you pick which AI model writes it (Claude or Gemini), with a credit estimate next to each. Either way the text only lands in the box: you still post it yourself.
Edit before posting
A drafted comment is a starting point. The value is that you were shown a relevant conversation with something to say, not that a machine wrote the words.
The scan reads publisher RSS and Atom feeds, and subreddits either through Reddit's official Data API (an app-level credential, never a connected account) or, without one configured, Reddit's native public per-subreddit feed. It uses an honest user agent, waits between requests, and backs off when Reddit's own rate-limit headers say to.
It never logs into anything or scrapes behind a login. LinkedIn and Instagram are never read this way.
Each item gets a score from your keywords, your persona, and how well the topic fits. Set a minimum per source so a busy subreddit does not fill the queue with noise.
Comments
Drafted replies to other people's posts, for you to review and post by hand. Studio, then Comments.
Postlyra never posts these
Copy and paste, always
Nothing on this page talks to any platform. Postlyra does not post comments, anywhere, ever.
The page hands you a draft and a link. You open the link, edit the comment, paste it, and record what you did.
Why: automated commenting is spam, it is against every platform's terms, and it is the fastest way to lose an account. Postlyra will not do it.
Where the drafts come from
Every feed and subreddit anyone tracks is read once, on a shared round-robin loop, whatever your configured Sources subscribe to. A busy platform round-trips faster; the schedule never depends on how many brands are watching the same feed.
New items get scored against your keywords and persona, and a comment is drafted in your voice for the ones worth replying to, into your queue.
Use Scan now on Sources to skip the wait for a feed you just added.
Working the queue
Three tabs: waiting, posted, skipped.
For each item you get the source post, a link, a relevance score, and a drafted comment.
Filter by source, minimum score, and date. Sort by newest, oldest or highest score.
Draft and Polish
The reply box is the same one the Inbox uses. Draft writes a comment from a direction you give it, Polish cleans up one you typed, and the Draft menu lets you pick which AI model writes it (Claude or Gemini), with a credit estimate next to each. Either way the text only lands in the box: you still post it yourself.
Edit before posting
A drafted comment is a starting point. The value is that you were shown a relevant conversation with something to say, not that a machine wrote the words.
Posting AI text verbatim reads as AI text.
Only public sources
The scan reads publisher RSS and Atom feeds, and subreddits either through Reddit's official Data API (an app-level credential, never a connected account) or, without one configured, Reddit's native public per-subreddit feed. It uses an honest user agent, waits between requests, and backs off when Reddit's own rate-limit headers say to.
It never logs into anything or scrapes behind a login. LinkedIn and Instagram are never read this way.
Scoring
Each item gets a score from your keywords, your persona, and how well the topic fits. Set a minimum per source so a busy subreddit does not fill the queue with noise.
Nothing in the queue