v0.5.0 · Open Source

Catch AI Trends
(Before Keyword Tools Do)

A daily radar for new AI model names, products and repositories — spotted on X, Hugging Face, GitHub and sitemaps before they appear in any keyword planner.

Get Started View on GitHub
Radar Pipeline Six signals → Name Judge → Keyword Table → Daily Report ① X Posts First mentions ① Hugging Face Trending models ② GitHub Breakout Viral repos ③ Sitemaps + New Sites New pages & domains Name Judge (LLM) Name vs common word Keyword Table Names + Generic terms, sized Daily Report Markdown + JSON output 14-Day Review Google Trends check ⑥ Revenue TrustMRR · Weekly ⑦ Reddit Asks What people request · Weekly ④ Search Console Your own fresh queries Use Cases Generic terms CLI & MCP Server python -m radar daily · Claude Code skill · MCP transport Cost: ~$15/month with example config (35 X accounts, 3x/day, keyword table for 5 names) With fewer accounts or once-daily reads, cost drops proportionally. NEW in v0.4: Revenue signal · Reddit asks · Page-one check · ⚡ marks

Six Discovery Signals

Each signal catches a different stage of a new term's emergence — from first mention to ranking pages.

SIGNAL ①

X + Hugging Face

Monitors your follow list on X and Hugging Face trending. Catches the very first mentions of a new name.

Free (HF) / API Every 8h
SIGNAL ②

GitHub Breakout

Finds repos under 14 days old gaining 200+ stars/day or 400+ in a single day. Adds search variants people use.

Free
SIGNAL ③

Sitemaps + New Sites

Watches AI tool directories for newly added pages, and finds domains registered in the last 90 days already ranking.

API Weekly
SIGNAL ⑥

Revenue

From TrustMRR: products people pay for. Fastest growing ($500+/30d) and newest listings — one at $0 earning now is flagged first.

TrustMRR Key Weekly
SIGNAL ⑦

Reddit Asks

12 phrasings of “is there an AI tool that…” A model keeps requests and complaints an AI tool could answer, written as people would search them.

API Weekly
SIGNAL ④⑤

Search Console + Review

Queries your own site got impressions for (fresh data). Then 14 days later each lead is checked in Google Trends against its baseline.

Free

The Keyword Table

A name is half of what you need. The table pairs each new name with the generic search terms behind it, sized and scored.

Keyword Type Est. Searches/Day Peak vs gpts Now Decision
photocraft 🎨 name 12,747 97 vs 3.1 still rising ✓ prioritize
claude haiku 5.5 name 1,863 100 vs 23.7 still rising ✓ prioritize
ai image detector 🔎 generic 2,537 87 vs 31.6 flat ⚠ watch
photoshop alternative ⚠ generic 244 61 vs 61.7 flat ✗ skip

Quick Start

Three ways to use the radar: CLI, MCP Server, or GitHub Actions.

Command Line

# Clone and run git clone https://github.com/VastFuture/new-term-radar cd new-term-radar python -m radar daily --out report.md
  • Install Python 3.10+
  • Set AISA_API_KEY environment variable
  • Copy config.example.json to config.json
  • Run python -m radar daily

MCP Server

# Install from PyPI pip install new-term-radar[mcp] # Or with uvx (no install needed) uvx --with mcp new-term-radar-mcp

Connects to Claude Code, Codex, or any MCP client. Passes AISA_API_KEY as environment variable.

GitHub Actions

# Set repo variables & secrets: # RADAR_ENABLED = true # AISA_API_KEY = your-key # Push config.json (remove from .gitignore first) git add config.json git commit -m "Add radar config"

Sources run every 8h. Daily runs everything including the keyword table. Reports appear in workflow summaries.

Ready to Catch the Next Big AI Term?

Star the repo, fork it to your org, and start running daily.

View on GitHub Read the Docs