OpenAI· Updated
OpenAI's Responses API now runs agents inside hosted containers with file systems, network access, and a shell tool. The agent loop handles long-running workflows by proposing shell commands, streaming results back, and compacting context automatically when the window fills.
grep, curl, awk; runs Go, Java, Node — not Python-only), hosted containers with file systems and SQLite, network access via an egress proxy with domain-scoped secret injection, and native context compaction via /compact. The API orchestrates the full loop — model proposes shell commands, container executes them, output streams back to the next step. Models GPT-5.2 and later are trained for this. Agent Skills — versioned SKILL.md bundles registered via API — load deterministically into container context.This removes the need to build custom execution infrastructure. The egress proxy addresses a production blocker: giving agents internet access without exposing credentials or allowing unrestricted outbound traffic.
Use the shell tool and container file system to build multi-step workflows that fetch live data, run scripts, and produce structured artifacts — the Responses API handles orchestration and compaction.
📣 Technical lessons from building computer access for agents Making long-running workflows practical required tightening the execution loop, providing rich context via file systems, and enabling network access with security guardrails. Here's how we equipped the Responses API with a computer environment: https://t.co/dMIlJN2iqA
Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →
