OpenAI releases Agents API in public beta, exposing Codex harness to developers
Following·Day 2·first covered 10 Sep 2026·2 sources
OpenAI releases Agents API in public beta, based on Codex harness
The gist
The API gives developers direct access to the same agent infrastructure OpenAI uses internally, removing the need to build custom context management and subagent coordination logic. Pricing carries no platform fee beyond token and tool costs.
The full picture
OpenAI released the Agents API in public beta on September 10, 2026, giving developers access to the same infrastructure that powers Codex and ChatGPT agents. The API handles context management, tool use, and subagent coordination through a single interface. Developers can run agents on an OpenAI-managed sandbox, their own infrastructure, or through ecosystem partners including Cloudflare, DigitalOcean, Vercel, and Modal. The API includes automatic context compaction for agents running for hours across multiple context windows, tool search that dynamically loads relevant tool definitions to reduce token usage, and multi-agent support that allows complex tasks to be broken into parallel subagents coordinated by a main agent. OpenAI charges no additional fees beyond token and tool usage. The underlying Codex harness is open-source.
How it developed
Sources
OpenAI Blog
The daily email
Want this in your inbox?
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free