OpenAI launches Agents API: sessions, orchestration and evals for hour-long agents
shaunralston · x · 2026-09-11
- OpenAI has published full documentation for its new Agents API, which the author calls a new engine for agents: the model was never the hard part — the hard part is an AI that can work for hours, use tools, remember what it did, recover from failures, and not choke on a mountain of logs.
- The API covers session management (create/resume), events and items, background mode, multi-agent orchestration, mid-turn steering, file inputs, compaction, and webhooks.
- It ships alongside Agent Builder (with a migration guide), agent safety guidance, and an Evals stack (prompt optimizer, graders, external models), plus fine-tuning options (SFT, DPO, RFT).
- Docs also reference newer models (e.g. GPT-6 Astra quickstart), realtime voice and voice agents, video generation, and a migration path from the legacy Assistants API.
Related event: OpenAI Launches Agents API Beta, Opening Up the Codex Harness(21 posts)→
More from coding & agent
- Stanford's PaperFormalizer turns research papers into Lean-verified versions with author in the loop — StanfordUniversity · 2026-10-27
- Hugging Face TRL makes score centering + masked importance sampling default in async GRPO — _lewtun · 2026-10-27
- AutoResearch agent's weird training curve gets dug into instead of discarded — ZimingLiu11 · 2026-10-27
- Firecrawl offers 2.5x credits to a16z Speedrun founders, unveils $75M Series B — devdigest · 2026-10-27
- Chums: open-source Apple Watch app puts Claude Code and Codex on your wrist — brandon_galang · 2026-10-27
- Open-source Clairic Studio turns pasted text into typed data and working code in 5 languages — dfinke · 2026-10-27