Show HN: Geiger – See every AI agent on your machine and what it can touch
Source Entity
Hacker News
Geiger is a new security diagnostic tool designed to inventory and audit AI agents and plugins running on a local machine. It addresses the growing risks associated with the rapid, unmonitored proliferation of autonomous AI tools.
The Rise of Silent AI Proliferation
The emergence of 'Geiger' marks a critical response to the chaotic expansion of the AI agent ecosystem. As developers and office workers alike integrate sophisticated AI harnesses and plugins into their local environments, the lack of transparency regarding what these programs can access has become a significant security liability. Geiger functions as a diagnostic tool, providing a read-only audit of every AI agent, plugin, and MCP server currently operating on a machine, effectively acting as a 'Geiger counter' for digital autonomy.
The Security Implications of Rapid Adoption
In August 2026, the technology landscape witnessed an unprecedented surge in the adoption of open-source agent harnesses, which garnered over 200,000 GitHub stars in just three weeks. This rapid scaling was accompanied by a massive explosion of the plugin ecosystem, which grew to over 13,000 repositories in that same window. This growth was fueled by viral social media content, specifically Instagram carousels that simplified the installation process for non-technical office workers, leading to widespread, often unvetted, deployment of powerful AI tools.
Understanding the Hidden Risks
Every AI agent installed during this period possesses the capability to execute system commands, read sensitive files, and hold user credentials—often configured within local dotfiles. Because these agents are frequently installed via one-click desktop clients, users are often unaware of the breadth of access they have granted. Geiger addresses this by scanning configurations and directories to provide a plain-language summary of exactly what each agent can 'touch,' bridging the gap between convenience and security.
Privacy-First Diagnostic Design
One of the most notable aspects of the Geiger tool is its commitment to privacy; it requires no installation, no account creation, and performs no telemetry. By operating as a read-only utility via npx, it ensures that the act of auditing one's machine does not introduce new vulnerabilities or data leaks. Users have the option to export the findings to a JSON file, but the tool itself writes nothing to the system, aligning with the growing demand for local-first, trustless security software.
Future Trends in Agent Governance
As AI agents become deeply embedded in local workflows, the future of cybersecurity will likely shift toward 'agent visibility.' Tools like Geiger represent the first wave of governance that empowers end-users to manage the blast radius of their AI software. Moving forward, we should expect to see more integration between diagnostic tools like this and OS-level permission frameworks, as the current model of installing powerful agents without granular oversight is increasingly viewed as unsustainable for enterprise and personal security alike.