Description
By the time a keyword tool lists a name like Seedance or Strata, you are about two weeks late and someone else's page already ranks. new-term-radar flips that: it camps where new names show up first, namely names the accounts you follow use for the first time on X, new models and Spaces on Hugging Face's trending list, GitHub repos gaining hundreds of stars a day, and new pages in the sitemaps of AI tool directories.
Every day it writes a keyword table: the new name itself plus the generic terms people will actually search for the thing it does, each sized, checked for room on page one, and given a decision. Which terms deserve a page first is spelled out in the table, so anyone running an AI tool site, a directory or a newsletter can read one report a day.
Many signals for new names: X, Hugging Face, GitHub breakout repos, eight AI directory sitemaps, new sites and payment links, each run on its own so they cross-check each other.
Name or common word: a small model scores every term as the name of one specific thing or an ordinary word, so everyday vocabulary does not flood the list.
Keyword table with a decision: estimated daily searches, when it started rising, when it was first searched, volume and difficulty across several markets, ending in prioritize, watch or skip.
Room on page one: it reads how strong the sites on Google's first page are and how the results are shaped, downgrades crowded terms, and tells you whether to update a page you already have or write a new one.
14-day check: each lead is compared against Trends two weeks later, giving a hit rate and lead time per signal.
CLI, Agent Skills and MCP: run it from the terminal, or plug it into Claude Code and similar tools as three Agent Skills or one MCP server. MIT licensed, Python 3.10+, and some signals need your own API keys, billed per call.
Every day it writes a keyword table: the new name itself plus the generic terms people will actually search for the thing it does, each sized, checked for room on page one, and given a decision. Which terms deserve a page first is spelled out in the table, so anyone running an AI tool site, a directory or a newsletter can read one report a day.
Features
Many signals for new names: X, Hugging Face, GitHub breakout repos, eight AI directory sitemaps, new sites and payment links, each run on its own so they cross-check each other.
Name or common word: a small model scores every term as the name of one specific thing or an ordinary word, so everyday vocabulary does not flood the list.
Keyword table with a decision: estimated daily searches, when it started rising, when it was first searched, volume and difficulty across several markets, ending in prioritize, watch or skip.
Room on page one: it reads how strong the sites on Google's first page are and how the results are shaped, downgrades crowded terms, and tells you whether to update a page you already have or write a new one.
14-day check: each lead is compared against Trends two weeks later, giving a hit rate and lead time per signal.
CLI, Agent Skills and MCP: run it from the terminal, or plug it into Claude Code and similar tools as three Agent Skills or one MCP server. MIT licensed, Python 3.10+, and some signals need your own API keys, billed per call.
