OpenAI announced on September 10 that it has released the Agents API in public beta, allowing developers to build and execute AI agents in the cloud. This API provides developers with access to the "Codex Harness," the agent execution framework operated internally by OpenAI.

With the Agents API, developers can build production-ready AI agents by specifying tasks, models, tools, and execution environments. For the execution environment, users can choose from OpenAI-managed sandboxes, services from partner companies such as Cloudflare, DigitalOcean, and Vercel, or their own Virtual Private Cloud (VPC).

In terms of functionality, the API features automatic context compression and support for tool connectivity via MCP (Model Context Protocol). Additionally, it offers multi-agent capabilities that allow complex tasks to be split into independent operations and assigned to multiple sub-agents for parallel processing.

The public beta is available to all developers. There is no additional fee for using the Agents API itself; charges are applied based on the token usage of the models and the tools employed.

Source: