OpenAI launches Agents API in public beta
OpenAI has released the Agents API in public beta, offering a managed backend infrastructure built on an open-source version of its Codex harness to power production-ready autonomous workflows. Designed to eliminate the operational overhead of custom agent scaffolding, the API natively manages workflow orchestration, persistent stateful sessions, automatic context compaction, error recovery, and sandboxed code execution in hosted or hybrid environments. The launch is paired with an open-source Agents SDK for Python and TypeScript, includes first-class support for Model Context Protocol (MCP) integrations and multi-agent delegation, and incurs no platform overhead fees beyond standard token and tool usage during beta.
OpenAI is transforming agent orchestration into a commodity cloud primitive, directly threatening standalone agent frameworks and hosting platforms while anchoring developers deeper into its proprietary ecosystem.
- –Commoditizes agent plumbing: Offloading persistent session management, context compaction, and failure recovery directly to OpenAI's managed runtime makes building custom agent loops largely redundant.
- –Normalizes MCP as the tooling standard: First-class integration with the Model Context Protocol cements MCP as the dominant standard for enterprise tool calling and agent interoperability.
- –Bridges flexibility and control: Developers can choose between the managed Agents API runtime or the granular, low-level control of the Responses API depending on their execution requirements.
- –Accelerates enterprise multi-agent architectures: Built-in agent-to-agent delegation and sandboxed runtime environments significantly lower the barriers to deploying autonomous agents in production settings.
DISCOVERED
1h ago
2026-09-10
PUBLISHED
1h ago
2026-09-10
RELEVANCE
AUTHOR
OpenAI