Introducing the Agents API

September 10, 2026 | Source: OpenAI Blog

Tags: OpenAI, Agents API, Codex, agentic AI, developer tools, cloud agents, orchestration

OpenAI launched the Agents API, a managed cloud service built on the Codex harness that lets developers create and deploy long-running AI agents with tool use and orchestration — entering direct competition with AWS Bedrock Agents and Google Vertex AI Agent Builder.

Details

OpenAI has introduced the Agents API, a new managed service for building and deploying cloud-based AI agents. The API is powered by OpenAI's Codex harness — the orchestration infrastructure behind their code-centric AI systems — and supports long-running sessions, multi-step tool use, and agent orchestration natively on OpenAI's infrastructure. This positions OpenAI directly in the agent infrastructure market, competing with Anthropic's API (tool use + computer use), AWS Bedrock Agents, Google Vertex AI Agent Builder, and emerging players like Langchain and Letta. A managed service means OpenAI handles session state, tool routing, and the orchestration layer, reducing infrastructure burden for developers building agent-based products. The Codex harness powering this API suggests agents will be particularly well-suited for software development, code execution, and structured technical workflows. Long-running session support addresses one of the main practical limits of current LLM APIs — tasks that require multi-step reasoning over extended periods without losing state. The source content is brief and does not disclose pricing, rate limits, access tiers, or current availability (GA vs preview), so the full commercial scope of the offering remains unclear from this announcement.