New Term Radar

new-term-radar is an open-source tool by Jason Zhu (zhuyansen) that catches the names starting to surge in AI, new models, products, tools and repos, where they are said first, before keyword tools list them. It then sizes the generic search terms behind each name into a keyword table with a decision. This pack bundles its three Agent Skills alongside a local MCP server and command-line tool. What it reads - X accounts you follow and Hugging Face's trending models and Spaces - GitHub breakout repos (young repos averaging hundreds of stars a day) and the search variants people use for them - Sitemaps of the AI tool directories you list, and newly registered sites that already rank - Reddit asks, new sellers from the public payment-link graph, and TrustMRR's verified revenue - Your own Google Search Console impressions, with DataForSEO Trends used only to check the radar against itself The keyword table - For the day's new names, it finds the generic terms behind them, sizes each one, checks whether there's room on page one, and gives a decision - A small model judges whether each term names one specific thing or is a common word How it installs - Install from the maker's Claude plugin marketplace, add the local MCP server, or use the Python package (python -m radar daily). It needs Python 3.10+. - Signals use paid data APIs: an AIsa key for X, new sites, Reddit asks, sellers and keyword sizing; a Google token for Search Console; optional GoAnyAPI and TrustMRR keys. Hugging Face and GitHub breakout signals are free. Good to know - The example config costs about $15 a month in API usage; costs scale with how many accounts and names you track. - MIT license.