OpenAI Releases Agents API Public Beta for Long-Running Tasks

01 What happened

On September 10, 2026, OpenAI introduced the Agents API in public beta, providing a managed harness for building and executing long-running autonomous agents.

02 Key details

• The API supports durable sessions, context compaction, tool searching, MCP integration, and custom function execution.
• Agents can be executed within OpenAI-hosted sandboxes, on external infrastructure, or via supported partner environments.
• OpenAI stated that the beta phase includes no additional fees beyond standard costs for model tokens and utilized tools.

03 Why it matters

This API centralizes session orchestration and execution, enabling developers to build persistent agents capable of handling complex, long-term workflows.

04 Who it matters to

Software developers and engineers building autonomous agentic applications and long-running AI workflows.