OpenAI Launches the Agents API in Public Beta, Putting the Codex Harness Behind One API Call

September 10, 2026 | Source: MarkTechPost

Tags: OpenAI, Agents API, Codex, MCP, multi-agent, developer tools, agentic AI

OpenAI opened its Agents API to all developers in public beta, providing the same managed harness and sandboxed infrastructure used to run Codex — including multi-agent coordination, durable sessions, and partner sandbox integrations from Cloudflare, E2B, and Vercel — at no extra fee beyond model costs.

Details

OpenAI has released the Agents API in public beta, giving all developers access to the managed infrastructure behind its Codex product. The API organizes around four concepts: Agent (model, instructions, tools, and MCP servers), Environment (an optional sandbox), Session (a durable agent instance), and Events/Items (inputs and outputs). Sessions can run for days, pausing and resuming across turns. Agents can run in three sandbox environments: an OpenAI-hosted sandbox (the same infrastructure behind Codex and ChatGPT for Work), a self-hosted option via codex exec-server, or partner sandboxes from Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop, and Vercel with first-class integrations. The announcement example shows a full incident-investigation agent — with multi-agent subagent delegation and observability MCP integration — created in a single API call. The harness handles context compaction, tool routing, and concurrent subagent coordination internally. Key constraints in public beta: data stays US-only and Zero Data Retention is not supported, which will block adoption for EU-regulated workloads. No additional API pricing is mentioned beyond underlying model costs.