<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> Agent Beacon addresses the fragmentation of AI agent activity across diverse runtimes by providing an open-source telemetry layer. It offers unified visibility for Security and IT teams, developers, and researchers, enabling consistent monitoring, security analysis, and debugging of AI agents wherever they operate—locally, in CI, or in the cloud.

How It Works

The project extends the OpenTelemetry GenAI standard, normalizing agent runtime events into a unified data model. Its architecture features local endpoint collection and processing by default, ensuring data privacy and control. Events are captured via hooks, wrappers, and SDKs at the agent runtime layer, processed locally for normalization and policy application at the endpoint layer, and then made available for inspection via a local dashboard or forwarded to external systems via the output layer. This approach provides a consistent event model across disparate agent surfaces while prioritizing local data handling.

Quick Start & Requirements

Installation on macOS is supported via Homebrew (brew tap asymptote-labs/tap && brew install beacon). Alternatively, users can build from source by navigating to cli/beacon and running make build. Official documentation, quick-start guides, and community resources are available for setup, deployment, and integration details.

Highlighted Details

• Extensive Runtime Support: Integrates with a wide array of local, CI, and cloud AI agent harnesses including Devin, Claude Code, GitHub Copilot CLI, VS Code, Anthropic, and OpenAI.
• Flexible Output Destinations: Telemetry can be forwarded to major SIEMs (Splunk, Sentinel, LogScale), log aggregators (Datadog, Elastic), and object storage (AWS S3, GCS).
• MDM Deployment Ready: Designed for enterprise rollout through standard MDM platforms like Fleet and Jamf Pro, facilitating security and IT team adoption.
• Local Dashboard & Threat Detection: Features a local, read-only dashboard for immediate log inspection and an offline threat detection engine using configurable rules.

Maintenance & Community

Agent Beacon is developed by Asymptote Labs. Community support and discussion are available via a linked Discord server. Specific details on core maintainers, sponsorships, or roadmap are not detailed in the provided README excerpt.

Licensing & Compatibility

The project is released under the permissive MIT license, generally allowing for commercial use and integration into closed-source projects without copyleft obligations.

Limitations & Caveats

The project emphasizes a "local-only posture" for security reviews, indicating that data is not sent externally by default, which requires explicit configuration for cloud-based aggregation. Some integrations depend on the specific telemetry exposed by the underlying agent harnesses, and the Vercel AI SDK integration is marked as experimental.