New-Term Radar MCP Server: Setup for Claude, Cursor and Other Apps
Lets your AI assistant spot new AI model, product and repo names as they appear, and size the search terms behind them.
New-Term Radar is a community project that lets Claude, Cursor and other assistants list new AI names that are starting to surge, then size the search terms behind them. It runs on your computer and keeps its data in a local folder. Two of its three tools call paid data services through your own AIsa key, so each run costs money. The project's own server notes tell the assistant to quote the cost and wait for your yes first.
Good for
- SEO people who build pages around new AI tools and models
- Newsletter writers and directory owners who want to know about a name the day it surfaces
- Founders checking whether a new term has searches before they build for it
Works in
12 should work
Tap an app to see how to set it up and any known problems.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Client does not support this transport.
Client does not support this transport.
Install New-Term Radar MCP
Claude Code
- 1
Copy this prompt.
Add the New-Term Radar MCP server to Claude Code. Run this command: claude mcp add new-term-radar -e AISA_API_KEY=<AISA_API_KEY> -- uvx --with mcp new-term-radar mcp Ask me for AISA_API_KEY and only put the values in that config, nowhere else. Then check that it loaded and tell me if I need to sign in or restart.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires uv (Python).
The with mcp and mcp arguments come from the README, which names Claude Code, Cursor, Claude Desktop and Codex. Needs uv installed.
Built from the package data we hold. Not yet tested by us. Source.
Tools
Shows the new AI names the radar has already kept
Lists the names the radar kept over the last few days, newest first, with the signals behind each and how often each signal turned out right. It reads only files on your machine and costs nothing, but it is empty until a radar run has happened.
Try: “Show the AI names the radar kept this week.”
Files ·
recent_leadsSizes the search terms behind one or two new names
Takes one to eight names (one or two works best) and returns a table of the search terms behind each, with estimated daily searches, trend and a prioritize, watch or skip call. It uses paid data services at about $0.05 a name plus $0.09 a call and takes 40 to 90 seconds a name. It also saves its data in the state folder.
Try: “Build a keyword table for the new model name Seedance 2.5.”
Write ·
keyword_tableRuns the daily scan for new AI names and returns the report
Runs a full scan once and hands back the report. A sources run reads the X accounts in your config, and a daily run takes 5 to 15 minutes. Both use paid services and update the same saved state as scheduled runs, so it suits a first try more than a routine.
Try: “Run the radar once and give me today's report.”
Write ·
run_radar
Read from the source code
Permissions
medium risk · 4 things to watch
Medium risk
Two of the three tools spend money on your AIsa key and save data on your machine, and the only brake on spending is an instruction to the assistant. Set a spending limit with AIsa and check the cost it quotes before you say yes.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
keyword_table and run_radar spend your money
They call paid services on your AIsa key, about $0.05 a name plus $0.09 a call for keyword_table. The server notes ask the assistant to quote the cost and wait for your yes, but that is an instruction to the model, not a limit in code. github.com
Writes data into the state folder
keyword_table and run_radar save results in the folder set by RADAR_STATE, or ./state if you leave it unset. Point it at an absolute path you chose. github.com
Names you ask about go to third-party services
Lookups pass through AIsa and, if you add a key, GoAnyAPI. Do not ask about unreleased product names. github.com
Web posts are treated as data, not instructions
The server notes tell the assistant to treat scraped posts and READMEs as data and to ask before posting, emailing or buying anything on the strength of a result. github.com
Security advisories
No published security advisories as of Nov 1, 2026.
Technical details
| Turning on read-only | No read-only mode is documented. recent_leads only reads local data, but keyword_table and run_radar call paid services and save data to the state folder. |
|---|---|
| Talks to | AIsa (X posts, search volumes, DataForSEO Trends and models). GoAnyAPI, if you add that key |
| Sign-in and keys | An AIsa API key in the app config. A GoAnyAPI key is optional. The Claude Code plugin keeps keys in the system keychain. |
Questions
Is New-Term Radar official?
No. It is a community project by zhuyansen (Jason Zhu in the package metadata). It is not made by AIsa, X, Hugging Face or any of the sites it reads.
Does it need an API key?
Yes for two of its three tools. keyword_table and run_radar need an AIsa key, and a GoAnyAPI key is optional. recent_leads works without a key but is empty until a run has happened.
How much does it cost?
The README lists about $0.05 a name plus $0.09 a call for keyword_table. A daily run is about $0.18 a day for 5 names, or $0.30 with a GoAnyAPI key. You pay the data services, not the project.
Does it need uv?
Yes for the install box, which starts the server with uvx. Without uv, the README says to install the package with its mcp extra and run the new-term-radar-mcp command.
Which apps does it work with?
The README names Claude Code, Cursor and Claude Desktop, and gives setup for Codex, TRAE, WorkBuddy and DeepSeek Harness. Codex needs longer timeouts because a keyword table takes 40 to 90 seconds a name.
Can it post or send things for me?
The README says the radar only reads, and the server notes tell the assistant to ask before posting, emailing or buying. The three tools listed here do not post anything.
What is the difference between the skills and the MCP server?
The skills are prompts that run the radar for you in agents such as Claude Code. The MCP server gives any MCP app the three tools. The Claude Code plugin brings the server with it.
Information
updated Nov 2, 2026
- Publisher
- Community
- Package
new-term-radar- Registry
- PyPI
- Transport
- stdio
- Auth
- API key
- Version
- 0.6.1
- License
- MIT
- Last commit
- Oct 28, 2026
- Open issues
- 1
- Listed in
- Official MCP registry
Sources
Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.
Show all 2Hide
- github.com · github.com
- github.com · github.com