MLSys 2026 Artifact Release

This release is an archival snapshot of the OpenHands Software Agent SDK created for the MLSys 2026 paper:

"The OpenHands Software Agent SDK: A Composable and Extensible Foundation for Production Agents"

This snapshot is intended for artifact evaluation and reproducibility.

What this release includes

• SDK implementation used in the paper
• Scripts for reproducing key experiments
• Event-sourcing traces and evaluation utilities
• Integration with benchmark evaluation pipelines

Notes for evaluators

• For full benchmark reproduction (e.g., SWE-Bench, GAIA), Docker is required due to sandboxed execution.
• Core functionality, SDK behavior, and event-sourcing measurements can be reproduced without Docker.
• Evaluation logs are available via:
  https://github.com/OpenHands/openhands-index-results

Important

This release is a frozen snapshot for reproducibility purposes and may differ from the latest development version on the main branch.

For the most up-to-date version of the project, please refer to the main repository.