HN Unveils Geiger to Show All AI Agents on Your Device and Their Access
agents
| Source: HN | Original article
Geiger, a new read‑only command‑line tool, inventories every AI agent, harness, MCP server, plugin and extension on a machine, showing what each can access.
A new open‑source utility called **Geiger** has been posted on GitHub, offering a single read‑only command that inventories every AI agent, harness, MCP server, plugin and AI extension installed on a computer. The tool scans known configuration locations and lists the components it can detect, warning users that even a single hit may indicate the agents are being used incorrectly.
Geiger’s approach is deliberately limited: it only reads configuration files, not runtime behavior, and it does not probe non‑standard install paths, other user accounts, containers or Windows Subsystem for Linux environments. Consequently, it cannot reveal what a plugin actually does, only where it resides and what it is permitted to access. The developers stress that the utility is a “Geiger counter” for AI agents – a diagnostic rather than a comprehensive security scanner.
The release matters because the rapid proliferation of locally‑run LLM agents, model‑hosting servers and third‑party extensions has outpaced visibility tools. As we reported on 10 September 2026 in our coverage of procedural graphs and multi‑agent prompt optimisation, developers are increasingly building complex agent ecosystems that can interact with files, APIs and system resources. Without a clear inventory, users risk unintended data exposure, privilege escalation or conflicts between overlapping agents.
Going forward, the community will be watching whether Geiger spurs broader adoption of inventory standards or inspires complementary tools that can monitor runtime activity. Security vendors may integrate its scanning logic into endpoint protection suites, while developers could extend its reach to containerised and WSL environments. The next wave of updates may also address the gap between configuration awareness and behavioural insight, offering a more complete picture of on‑device AI activity.
Sources
Back to AIPULSEN