We're proud and excited to be a first-class sandbox provider in the new OpenAI Agents SDK!
With this SDK, agents can directly create and deploy Blaxel sandboxes on demand to run commands, create files, and write and test arbitrary code. These sandboxes run on Blaxel’s high-performance, AI-native infrastructure platform, with automatic scale-to-zero and world-class boot/resume times.
This integration is perfect for building the next generation of coding agents or data analyst agents:
- Each agent gets an isolated Blaxel sandbox per task and can freely execute arbitrary code without any risk to the host.
- Blaxel sandboxes resume from standby in ~25ms with full memory state, producing unbeatable performance for almost all sandbox operations. You can choose to auto-delete them once the agent finishes, or keep them running for longer durations.
- Agents can render live application previews as they work, so human users can build, step away, and inspect or provide further instructions when ready. Both public and private previews are supported.
- The system scales horizontally without any manual coordination. Agents can work in a single sandbox or spin up multiple sandboxes per task.
Once you've built and tested your agents, you can also deploy them to Blaxel with minimal code editing and zero configuration, enabling you to benefit from near-instant response times by colocating them close to the sandboxes the agents are already using.
Get started with our integration guides here: https://docs.blaxel.ai/Tutorials/OpenAI-Agents-SDK-Index
