OpenAI Agents API

Managed API, in beta, that runs the Codex agent harness with sessions, sandboxes and tools.

Software you sign up for or install and run yourself. Submitted listing, not yet checked. Nothing on this page is sponsored.

What it does

The OpenAI Agents API gives an application access to the Codex harness as a service run by OpenAI, which manages sessions, orchestration, context compaction and recovery. A developer defines an agent with a model, instructions, tools and MCP servers, chooses an OpenAI-hosted sandbox or a self-hosted one, and sends user messages that start turns of work in which the agent runs commands, edits files and can delegate to subagents. Results arrive as streamed events or webhooks, along with artifacts produced in the sandbox. The endpoints sit in a beta namespace, usage is billed at the chosen model's API rates plus standard tool and container rates, data residency is United States only and Zero Data Retention is not supported.

Type
Software product
Pricing
Usage-based
Setup
Developer setup
Access
API
Human review
Runs without review
Vendor
OpenAI
Website
developers.openai.comChecked 19 October 2026
Listed
19 October 2026

You provide

  • An agent definition with model, instructions and tools
  • MCP servers
  • An execution environment, hosted or self-hosted
  • User messages that start a turn

You receive

  • Streamed events
  • Webhooks on completion or when input is needed
  • Files and artifacts from the sandbox

How the work runs

  1. 01

    Define the agent and its tools

  2. 02

    Choose a hosted or self-hosted sandbox

  3. 03

    Send a user message to start a session

  4. 04

    The agent runs commands, edits files and calls tools, keeping state across turns

Before it is handed over

Runs without review.

Find more like it

Weighing it up? Compare OpenAI Agents API with 10 alternatives side by side.

← All agents

Agents that share this one's category, use cases or setup.