Agents API

OpenAI's managed API bringing Codex's agent harness, durable sessions, and hosted sandboxes to developers building production agents.

OpenAI · Agent frameworks and orchestration · framework

Published in Feed

Overview

OpenAI introduced the Agents API in public beta on September 10, 2026, exposing the durable session and orchestration layer behind Codex to third-party developers. It manages context, recovery, and multi-step execution for agents that run continuously in production, with automatic context compaction, multi-agent orchestration, programmatic tool calling, and MCP server support. OpenAI hosts and manages the underlying sandbox by default, but developers can instead run agents on their own infrastructure or through supported sandbox providers including Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop, and Vercel. On September 29, OpenAI added computer use through an OpenAI-hosted browser, with website access approvals and sign-in handled by the developer's application. In the week of October 30, 2026, OpenAI engineers described further additions: Auto Review, which checks commands for bad behavior before they run; background async compaction; experimental time awareness for acting on time limits; mid-session tool updates; and automatic system-prompt caching across sessions. The official API changelog had not yet listed these when reported, so parameter names and availability are unconfirmed.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Durable, recoverable multi-step agent sessions with automatic context compaction
  • Multi-agent orchestration and programmatic tool calling
  • MCP server support
  • OpenAI-managed sandbox, self-hosted, or third-party sandbox provider (Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop, Vercel)
  • Computer use in an OpenAI-hosted browser, added September 29, 2026
  • Week of Oct 30, 2026: Auto Review, async compaction, experimental time awareness, mid-session tool updates, automatic system-prompt caching (per OpenAI engineer post; changelog not yet confirmed)

Use cases

  • Building production agents that run continuously without custom orchestration infrastructure
  • Enterprise agent development needing durable sessions and recovery
  • Agents that need a secure sandbox to run code, edit files, and produce artifacts

Pricing

No additional platform fee; billed only for the tokens and tools the agents use.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
10 Sept 2026
Record updated
1 Nov 2026
Record last checked
1 Nov 2026

Explore the Feed

Other entries filed under agent frameworks and orchestration; inclusion does not imply equivalent features.

  • Workday Agent Harness

    Workday's generally available environment for building enterprise agents with isolated runtimes, memory and governed HR/finance data access.

  • Agent Bricks

    Databricks' no-code system for building and auto-optimizing domain-specific AI agents, now generally available with synthetic eval-data generation and multi-agent orchestration.

  • Muse Agents SDK

    Meta's open-source SDK for building and orchestrating multi-agent systems on top of its Muse model family.

Browse the Feed