Three paths. Different control.

Choose where the agent loop runs and which infrastructure your application owns.

Where do you want control?

Agents API

Durable sessions, orchestration, and optional hosted compute.

Read about Agents API
OpenAI agent runtime comparison
ResponsibilityAgents APIAgents SDKResponses API
Agent loopOpenAI-managed harnessSDK runner in your runtimeYour application
StateDurable API sessionsYour session/storage integrationYour state or API conversations
Custom functionsYour application handlerYour application toolsYour application handler
Execution environmentHosted, self-hosted, or noneYour runtime and integrationsYour infrastructure
DeploymentIntegrate the managed serviceDeploy your agent applicationDeploy your orchestration
Best fitManaged, continuing agent workApplication runtime controlCustom model-and-tool workflows

These are different runtime choices. A session, sandbox, SDK runner, and model response are not interchangeable resources.

Official runtime comparison ↗
Read the full comparison →