axonpush

axonpush provides robust agent monitoring and testing capabilities, ensuring the reliability of complex agent systems. Key features include: • Replay production failures against new changes • Evaluate model calls and full request traces • Identify issues from tool errors to retrieval failures • Integrate with CLI and GitHub Actions for continuous validation This platform addresses the critical challenge of maintaining stable agent performance. When a prompt, a model, or an agent step is modified, axonpush automatically re-runs historical production failures. If any previously working scenarios break, the build fails, preventing regressions from reaching live environments. This goes beyond simple model evaluations by replaying the entire request lifecycle, capturing errors from external tools, empty data retrievals, and query timeouts. axonpush offers deep insights into agent behavior, allowing developers to pinpoint the exact cause of failures. The system provides clear feedback through an exit code 1, a JUnit file, and direct PR comments, streamlining the debugging process. It ensures that every change is thoroughly validated against real-world scenarios, improving stability and reducing operational risks. Built for engineering teams and developers working with sophisticated agent systems, axonpush helps maintain high-quality outputs. It ensures changes are safe to deploy by catching regressions early, making it an essential tool for continuous integration and delivery pipelines.