Sign in

A human-in-the-loop X tool

Replies worth sending.

Reply Radar watches a curated roster of X accounts, scores every fresh tweet for velocity and relevance, and drafts replies and original posts in your voice. You read, pick, and send. It never posts on its own.

113,967
tweets scored
163
accounts watched
98
human-approved sends
The Reply Radar feed: roster tweets ranked into tiers with velocity and relevance scores
The feed — every roster tweet from the last few days, ranked. Tier 01 is what deserves a reply today.
A tweet page with three drafted replies, each with revise, vote, and send-on-X controls
Drafts — candidates in the profile's voice. Revise with one instruction, vote to teach the guide, send from X.
The weekly engagement report: aggregates over sent replies and posts plus model findings
What worked — a Monday report over the week's sends. Recommendations promote straight into the voice guide.

How it works

  1. 01

    Fetch

    New tweets from a curated X list land every few hours over a read-only API.

  2. 02

    Classify

    A small model rates each tweet's relevance to the profile's niche.

  3. 03

    Score

    Velocity × relevance × roster tier × recency decay ranks the feed into tiers.

  4. 04

    Draft

    A frontier model writes reply and post candidates in the profile's voice; a judge pass throws out reply-guy slop.

  5. 05

    Send

    You pick one, edit if you like, and post it yourself from an X intent link.

  6. 06

    Learn

    Nightly engagement snapshots and a Monday “what worked” report feed back into the voice guide.

Built with

  • Next.js 16
  • Trigger.dev
  • Supabase Postgres
  • Anthropic Claude
  • TwitterAPI.io
  • Vercel