OpenAI launches Agents API to streamline autonomous agent development

OpenAI has launched its Agents API in public beta, a managed service designed to simplify the creation and deployment of autonomous AI agents. The API provides access to the underlying infrastructure and orchestration harness used by Codex and ChatGPT for Work, allowing developers to build custom agents with a single API call by specifying the task, model, and available tools.

To reduce manual engineering requirements, the service handles backend tasks such as session management, execution environments, and tool orchestration. Key features include automatic context compaction to manage session limits, intelligent tool searching to reduce token consumption, and programmatic tool calling for parallel or chained execution. The API also supports multi-agent workflows where a primary agent can decompose complex tasks.

Developers have flexible options for execution environments. Agents can run in an OpenAI-managed sandbox, on a company’s own internal infrastructure, or through third-party providers including Cloudflare, DigitalOcean, and Oracle. This flexibility allows enterprises to balance ease of use with specific security, compliance, and existing cloud relationship requirements.