OpenAI Agents SDK - Build production agents with harness and sandbox
The updated OpenAI Agents SDK introduces a model-native harness and native sandbox execution. Build agents that safely inspect files, run commands, and execute code for long-horizon tasks across built-in providers like E2B, Modal, Daytona, and Vercel.
Replies
Flowtica Scribe
Hi everyone!
@OpenAI ’s updated Agents SDK adds two big pieces for production agents: a model-native harness for long-horizon work across files and tools, and native sandbox execution (including @cloudflare, @Modal, @Vercel and @E2B) so agents can inspect files, run commands, edit code, and keep working safely in controlled environments.
The production agent infrastructure race is getting very real!
I'm particularly interested in how the harness handles long-horizon agent tasks; I often struggle with state tracking.
Open Wearables
finally, a harness that doesn't feel like an afterthought. been using Cursor and Claude Code daily and the safety guardrails are always the weakest link. love that you're tackling file inspection and command execution head-on instead of just wrapping the API calls.
native sandbox execution across E2B/Modal/Daytona/Vercel is a nice touch — does the SDK abstract away the provider differences or do you still pick one explicitly per run?