Suggestion mode by default. Auto mode when you trust it.
Start safe — every reply is a draft until you click Publish. Once you've seen enough good drafts, flip the switch and let ReplyMind post as a child comment automatically, attributed to the post author.
Eight models. One key.
Bring your own OpenAI or Anthropic key. Swap models from a dropdown — Sonnet 4.6 by default, or fall back to Haiku 4.5 when cost matters.
Author-aware
Replies post under the post author's name, email and URL — your readers feel a person answered, not a script.
Batch & bulk publish
Generate drafts for every existing approved comment in chunks of 2. Publish drafts in chunks of 5. Live progress bar, stop button.
Audit log
The last 200 generations, newest first. Success or error, full reply text, links to the comment and post. Cleared with one button.
Daily rate limit
100 successful generations per day, by default. Filterable. Failed calls don't consume quota. Counter resets at midnight.
Never blocks visitors
Comments are processed by WP-Cron, not synchronously. The visitor's submit POST returns instantly — no external API in the request path.
Prompt-injection guard
The system prompt explicitly instructs the model to treat the comment as untrusted data and ignore embedded instructions.
Inline draft modal — edit and publish without leaving the page
Click Edit & Publish on any draft row. The modal opens pre-filled. Tweak a sentence, click publish. AJAX posts the reply as a child comment. ESC, outside-click and Cancel all dismiss.
Filter to drafts
Add ?replymind_filter=drafts to the Comments URL — only the ones awaiting your review.