Book a call

News · September 10, 2026

OpenAI opens a managed agent runtime to developers in beta

// ai.news :: 2026

OpenAI's new Agents API offers a hosted version of the harness behind Codex, so a custom agent can be created with a single API call. It removes infrastructure work, but data retention and vendor lock-in deserve a close look before production use.

What happened

  • OpenAI announced the Agents API in public beta on September 10, 2026, open to all developers, with no extra fee beyond the tokens and tools the agents consume.
  • According to OpenAI, the hosted harness is based on the open-source Codex harness and is updated with each model launch. Agents can run in OpenAI-hosted sandboxes, on a company's own infrastructure, or with partners such as Cloudflare, Vercel, Modal and E2B.
  • OpenAI says the API compacts earlier context automatically in long sessions, loads tool definitions only when needed, supports MCP and custom functions, and lets a main agent split work across parallel subagents with a configurable limit.
  • OpenAI's documentation states that the Agents API does not support Zero Data Retention, even with a self-hosted sandbox, and that data residency is currently available only in the United States.
  • InfoWorld reported analyst concerns that relying on one vendor for the model, context handling, tools and execution makes switching harder, and named competing services from Anthropic, AWS and Microsoft, as well as LangGraph.

Why it matters

A managed harness removes a lot of plumbing. Sandboxing, context compaction and subagent orchestration are hard to build well, so for prototypes and internal tools this can noticeably shorten the path from idea to a working agent.

For production, the documentation matters more than the launch post. No Zero Data Retention and US-only data residency rule out some workloads, especially for European customers or regulated data. Check this against your data classification before committing.

To limit lock-in, keep your tools behind MCP or your own function layer, and keep evaluation sets independent of any one vendor's runtime. Then the agent logic can move if pricing, terms or quality change.

Sources

This post is CMM’s own summary and commentary on publicly reported news. Facts are attributed to the sources linked above; please read them for full details. CMM is not affiliated with these publishers or with the companies mentioned, and trademarks belong to their owners. Prepared with AI assistance. Corrections or requests: [email protected]. See our editorial policy.

Want help with this?

Tell us what you need, and we will suggest the right team model on a short call.

Talk to us