OpenAI Agents API
Featuredby OpenAI
Run OpenAI's Codex harness as a managed service — durable sessions that survive restarts, hosted or self-hosted sandboxes, MCP connections, vaults, and multi-agent delegation.
Skills
Durable Sessions
Runs long-lived agent sessions that persist across requests, so work can be paused, resumed, and continued later.
Sandbox Execution
Executes agent code in OpenAI-hosted or self-hosted sandboxes, with lifecycle and security controls per environment.
MCP Connections
Connects a session to external MCP servers and plugins so the agent calls your tools alongside built-in ones.
Session Webhooks
Emits webhook events for session progress and completion so long-running runs report back without polling.
Related Agents
CrewAI
Multi-agent orchestration framework for building teams of AI agents that collaborate, delegate, and solve complex tasks…
SandBase Harness
Runs local-first, self-hosted AI agents with persistent sessions, sandboxed tools, memory, credentials, audit trails, r…
A2A Protocol
Defines the Agent2Agent interoperability protocol and SDKs so agents built on different frameworks discover each other,…
AutoGen
Microsoft's open-source multi-agent framework. Supports conversational, autonomous, and human-in-the-loop agent pattern…