Geiger: An Open-Source Tool That Maps Every AI Agent on Your Machine and Its Reach

A developer going by Atomburstofficial has released Geiger, a new open-source tool designed to expose the AI agents operating on a local machine and map exactly what each one can touch. The project, posted to Hacker News today, aims to solve a growing visibility problem: as more developers run autonomous coding assistants, local LLM pipelines, and agentic scripts, it's easy to lose track of what those processes are doing and which parts of the system they can reach. Geiger scans running processes, identifies AI-driven agents, and builds a clear map of their permissions and potential impact zones.

The tool's name is a deliberate nod to the radiation detector—it's meant to give users a sense of 'ambient AI activity' that might otherwise go unnoticed. Instead of waiting for a security incident or a rogue agent to overwrite files, developers can now see, at a glance, which agents are active, what files they've accessed, and what system resources they can potentially modify. That kind of transparency is increasingly critical as AI agents move from experimental scripts to production workflows that handle real data.

Geiger arrives at a moment when the AI agent ecosystem is expanding rapidly, but tooling for oversight hasn't kept pace. Most existing security and monitoring tools treat AI agents like any other process, missing the specific risks they introduce—such as an agent with file-write access that could inadvertently corrupt a codebase, or one with network permissions that might exfiltrate sensitive data. By focusing specifically on agents and their 'touch' scope, Geiger offers a niche layer of introspection that general-purpose system monitors don't provide.

The initial Hacker News response has been modest—14 points and two comments at the time of writing—but the project taps into a real concern shared by many in the developer community. One commenter has already asked about compatibility with containerized environments, a sign that users are thinking about how to apply Geiger beyond a single machine. The creator appears open to feedback, sharing the project as a 'Show HN' to invite collaboration and feature requests.

For now, Geiger is a command-line tool that requires some technical comfort to install and run. It's written to work on common Unix-like systems, and the GitHub repository includes installation instructions and examples of the output. The project is still in its early stages, so expect rough edges, but the core premise—making AI agent activity visible and auditable—is one that will likely resonate with developers who are tired of flying blind.

Looking ahead, the bigger question is whether Geiger can evolve into a standard part of the AI development toolkit. For that to happen, it will need to support a wider range of agent frameworks, offer a more user-friendly interface, and perhaps integrate with existing observability platforms. The project's open-source nature means that progress depends on community interest, but if the demand for AI agent transparency continues to grow, Geiger could become a foundational utility for anyone running multiple agents on their machine.