OpenAI Opens Its Codex-Powered Agents API to All Developers in Public Beta
OpenAI's Agents API is now in public beta for all developers. The managed service uses the Codex harness to run long-lived cloud agents, coordinate parallel subagents, compact context across long sessions, and work in OpenAI-hosted, VPC, or partner-provided sandboxes.
OpenAI has opened the Agents API to all developers in public beta. The managed service brings the Codex harness to cloud agents, handling orchestration, long-running sessions, and context management while developers choose the task, model, tools, and execution environment. (OpenAI Developers on X; OpenAI's Agents API announcement)
The launch is aimed at teams that want to build production agents without first assembling their own agent loop. OpenAI says the service is fully managed, uses the same harness and infrastructure that power Codex, and lets developers focus on the tools, knowledge, and workflows that make an agent useful. (OpenAI's Agents API announcement)
What the Agents API handles
OpenAI describes the Agents API as a foundation for agents that can run reliably for days. Agent environments can run code, work with files, save intermediate results, and produce artifacts. The API is responsible for coordinating model calls, tools, and context through the Codex harness. (OpenAI's Agents API announcement)
The API also includes built-in support for longer workflows. OpenAI says automatic context compaction preserves the information an agent needs as a session approaches its context limit, allowing a workflow to span multiple context windows without developers implementing their own compaction logic. (OpenAI's Agents API announcement)
Parallel subagents and programmatic tools
The Agents API includes multi-agent support. OpenAI says a main agent can split a complex task into independent pieces, delegate them to subagents working in parallel, and combine their results. Each subagent keeps its own context, which can help with research, analysis, and coding tasks that benefit from parallel work. (OpenAI's Agents API announcement)
The API also includes tool search and programmatic tool calling. Tool search loads relevant tool definitions as needed, while programmatic tool calling lets an agent run calls in parallel, chain operations, filter results in code, and combine results before bringing the useful material back into model context. (OpenAI's Agents API announcement)
OpenAI says supported tool types include MCP, custom functions, and built-in tools such as web search. (OpenAI's Agents API announcement)
Choose the sandbox
The Agents API separates the managed agent harness from the place where the agent runs code and handles files. OpenAI says developers can choose among an OpenAI-managed sandbox, their own infrastructure including deployments inside a VPC, or a sandbox from a partner provider. (OpenAI's Agents API announcement)
OpenAI's hosted sandbox uses the same sandboxing infrastructure that powers Codex and ChatGPT. The company says it provisions and manages the environment, which can be configured with files, packages, skills, and plugins. Agents can run code, work with files, and produce artifacts inside it. (OpenAI's Agents API announcement)
For teams that need different infrastructure choices, OpenAI lists first-class partner integrations with Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop, and Vercel. OpenAI says those options can differ in file and secret storage, CPU and GPU configurations, memory, performance, cold starts, and cost profiles. (OpenAI's Agents API announcement)
What developers should expect in beta
OpenAI says the Agents API is available today to all developers in public beta. The company says there are no additional fees for using the API itself; users pay for the tokens and tools their agents use according to the linked pricing details. (OpenAI's Agents API announcement)
The public-beta status matters. OpenAI says it will iterate quickly based on feedback while working toward general availability, and the announcement does not provide a general-availability date or promise a fixed set of limits. (OpenAI's Agents API announcement)
How it differs from a basic model call
A basic model call returns an answer for a request. The Agents API is designed to keep a managed agent working across long sessions, coordinate tool calls and subagents, preserve relevant context, and run code or manipulate files in a chosen environment. That is the product boundary OpenAI is emphasizing: developers bring the task and application-specific tools, while OpenAI operates the harness and the cloud-agent infrastructure. (OpenAI's Agents API announcement)
The announcement does not establish a guaranteed completion time, universal tool availability, fixed sandbox quotas, or a specific model requirement. Those details should be checked against the current API documentation and pricing before production deployment. (OpenAI's Agents API announcement)
Bottom line
OpenAI's Agents API makes the Codex harness a managed building block for long-running cloud agents. Its public-beta feature set combines multi-agent orchestration, automatic context compaction, programmatic tool use, and a choice of OpenAI, VPC, or partner sandboxes. For developers, the practical change is that the agent runtime and its operational scaffolding can be selected rather than built from scratch. (OpenAI's Agents API announcement)
Sources
- OpenAI Developers on X, “Go from idea to a working agent faster with the Agents API,” September 10, 2026: https://x.com/OpenAIDevs/status/2098130570048045453
- OpenAI, “Introducing the Agents API”: https://openai.com/index/introducing-the-agents-api/
Read next
Keep building the workspace playbookOpenAI Releases Codex Security CLI for Repository Scans and CI Checks
OpenAI says its open-source Codex Security CLI can scan repositories, track findings across runs, verify fixes, and add security checks to CI/CD. The beta CLI requires Codex Security access and is built for teams that want code-aware security review in the terminal.
OpenAI's Secure MCP Tunnel: Connect Private MCP Servers Over Outbound-Only HTTPS
OpenAI's Secure MCP Tunnel connects private and on-prem MCP servers to ChatGPT, Codex, and the Responses API without opening inbound firewall ports. You need a tunnel_id, an API key with Tunnels Read plus Use, and -- separately -- ChatGPT developer-mode access. That permission split is the most common setup blocker.
Frequently Asked Questions
What is OpenAI's Agents API?
The Agents API is a managed OpenAI service for building and running cloud agents with the Codex harness. Developers specify a task, model, tools, and environment instead of building the agent harness and orchestration layer themselves.
What can Agents API sessions do?
OpenAI says agents can run for days, work with files, run code, save intermediate results, produce artifacts, and coordinate subagents working in parallel. The API also automatically compacts earlier context as a session approaches its limit.
Where can an OpenAI cloud agent run?
OpenAI says developers can choose an OpenAI-managed sandbox, their own infrastructure including a VPC deployment, or a sandbox from an OpenAI partner such as Cloudflare, Daytona, Modal, Runloop, or Vercel.
Is the Agents API generally available?
No. OpenAI says the Agents API is available to all developers in public beta. The company says it will iterate based on feedback while working toward general availability, but does not give a general-availability date in the announcement.