GraphOS – Visual runtime and debugger for AI agents (with local-first execution) (github.com)

GraphOS has launched as an open-source governance and observability layer for LangGraph.js, designed to enhance the management of AI agents in production. This tool addresses critical issues faced by AI developers, such as infinite loops, runaway costs, and the black-box phenomenon of complex AI runs. GraphOS ensures policy enforcement with features like LoopGuard and BudgetGuard, which monitor agent behavior in real-time, halting executions when necessary and preventing agents from exceeding predefined cost limits. Significantly, GraphOS allows developers to visualize the execution of AI agents through a local-first dashboard that supports time-travel replay of sessions. The integration of real-time monitoring with the capability to trace and interact with execution data increases transparency, making it easier for developers to debug and optimize AI workflows. With support for both TypeScript and Python, GraphOS's SDK facilitates seamless integration into existing applications, promoting a more rigorous and cost-effective deployment of AI agents without the need for external service dependencies or telemetry.