Skip to main content

Reply engine

Scale engagement without scaling the ban risk.

The MarquIQ reply engine scans conversations across Reddit, Hacker News, Indie Hackers, Dev.to, and Hashnode, drafts responses grounded in your product context, and routes everything through human approval on platforms where a mistake is unrecoverable.

Automated replies are a minefield. Platforms notice generic output faster than humans, and the penalty for getting it wrong is your account. The reply engine is built around the guardrails that separate signups from bans.

What you get

  • SKIP signal against off-topic replies

    The drafting model can refuse to draft. Threads that are not a fit are marked engaged but never produce a reply that could ban you.

  • Confidence thresholds per platform

    X and Bluesky auto-send above 0.8 confidence. LinkedIn above 0.85. Reddit, HN, Indie Hackers always require human approval regardless of confidence.

  • Per-subreddit and per-platform caps

    Two replies per subreddit per day. Ten Reddit replies per account per day. Fifty engagements per day globally. Caps are structural, not advisory.

  • Mobile approval flow

    PWA and Telegram bot let you clear the queue from a phone. Each approval card is under twenty seconds end-to-end.

  • Measurement loop built in

    Every reply scored at 6, 24, and 72 hours. After 30 days, the tuning engine surfaces keywords and subreddits to drop because they never convert.

  • Platform-native tone

    HN tone is technical and sparse. Reddit tone matches the subreddit. Indie Hackers tone is founder-to-founder. The engine does not shift one voice across surfaces.

How it works

  1. Step 1

    Define keywords and surfaces

    Pick subreddits, HN query terms, Indie Hackers categories, and the keywords that signal a buying-moment conversation.

  2. Step 2

    Scan conversations

    The crawler pulls fresh threads on a schedule. Every thread is scored for relevance before it enters the drafting queue.

  3. Step 3

    Draft or SKIP

    The drafting model either produces a draft grounded in your product facts or returns SKIP_NOT_RELEVANT. Off-topic replies never queue.

  4. Step 4

    Route by risk

    Low-risk surfaces route to auto-send above threshold. High-risk surfaces always route to human approval. The router is configurable per organization.

  5. Step 5

    Score and tune

    Every posted reply is scored at 6, 24, and 72 hours on upvotes, comments, link clicks, and attributed signups. The tuning engine surfaces what to stop.

Frequently asked questions

Does the reply engine auto-post to Reddit?

Never. Reddit, Hacker News, Indie Hackers, Dev.to, and Hashnode always require human approval, regardless of confidence. The downside is account-level; we refuse to take that risk on your behalf.

What is the SKIP signal?

The drafting LLM can return SKIP_NOT_RELEVANT instead of a draft. If the thread is not a fit, the engine marks the conversation as engaged but does not draft anything. This is the single most important guardrail against off-topic replies.

Will the engine detect platform rate limits?

Yes. Per-platform daily caps, per-subreddit caps, and cross-subreddit caps are enforced at the engine level. Cooldowns between replies on the same thread are two hours minimum.

How does approval work on mobile?

The PWA and Telegram bot show the thread summary, the draft, the confidence score, and one-tap approve, edit, or reject. Clearing the queue from a phone takes under twenty seconds per reply.

What happens if I get banned anyway?

The engine flags the account, pauses posting, and surfaces the ban for review. You can re-connect once the platform resolves. MarquIQ does not bypass platform bans or re-auth through alternate accounts.

Run the reply engine without ban risk.

Every guardrail in this page is a default. Set confidence thresholds, approve from mobile, and let the measurement loop narrow your surface area to what actually converts.

Start the trial

Related reading