Geiger Scans a PC and Lists Every AI Agent and MCP Server It Finds
An inventory tool answers a question that arrives only after people have installed a stack of agents and MCP servers and forgotten where each one sits, and Geiger's labels make the access each item already holds readable without exposing the credentials themselves.
Reporting from 1 source: GIGAZINE.
Geiger, a free tool published on GitHub by Atomburstofficial, inventories the AI agents, harnesses, MCP servers, Hooks, and AI extensions configured on a machine. It runs through npx on Node.js 18 or later, with no global install and no account. Each detected item is labeled for what it can reach, such as code execution, credentials, filesystem, or network access, and credential values are reported by key name only, not shown.
The scan command is a single line: npx geiger-scan. Its output groups agents such as Claude Code and Codex alongside MCP servers, Hooks, and extensions registered to Cursor and VS Code, so the list reflects what is on the machine rather than what one editor knows about.
Labels attached to each entry describe capability: EXECUTES, HOLDS-SECRETS, BROAD-FILESYSTEM, BROAD-WEB, NETWORK, UNKNOWN-ORIGIN. A --html option writes a self-contained report with each item's configuration file and remediation steps, and --json saves a baseline that --diff compares against a later scan. Where a configuration file holds something credential-shaped, Geiger reports the key name, the file, and the credential type, not the value.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.