Agent Beacon: Open-Source Telemetry for AI Agents Launched by Asymptote Labs

Asymptote Labs has officially released Agent Beacon, an innovative open-source telemetry layer specifically engineered for AI agents. This new tool addresses the growing need for transparency and control over the operations of AI coding and knowledge-work agents, such as Claude Code, Codex CLI, Cursor, and Claude Cowork. Agent Beacon works by configuring these AI runtimes to automatically export OpenTelemetry data, which is then directed to a collector running on the local machine. The core functionality of Agent Beacon involves taking the diverse telemetry streams from various AI agents and normalizing them into a consistent JSON event log. This unified log provides a clear, chronological record of agent activities, including sessions, prompt submissions, tool usage, command executions, and file modifications. A local dashboard is included to allow developers and security teams to inspect these activities in real-time. Beyond local inspection, Agent Beacon supports forwarding this rich telemetry data to various customer-managed SIEMs and log destinations, including Elastic, Microsoft Sentinel, AWS CloudWatch, and Google Cloud Storage. This integration capability is crucial for enterprises looking to incorporate AI agent activity into their broader security and operational monitoring frameworks. A key consideration in AI agent telemetry is data privacy and security, especially concerning sensitive information like prompt text. Agent Beacon tackles this by offering configurable retention modes. Users can choose to redact or limit the size of specific content fields, such as prompt text, raw attributes, command output, and raw diffs, before the data is logged or forwarded. This ensures that organizations can maintain necessary visibility while adhering to privacy and compliance requirements. Written primarily in Go and released under the MIT license, Agent Beacon is designed for ease of deployment, with macOS installation available via Homebrew. The project underscores a significant step towards establishing robust observability and governance practices for the rapidly evolving landscape of AI-driven development and operations. Asymptote Labs envisions Beacon as a foundational layer for understanding and securing AI agent behavior across local, CI, and cloud environments.