Create Agents API resources from your own program or terminal. POST /v1/agents saves a reusable configuration; POST /v1/agents/sessions starts a working session. The second request can include the configuration and first task, so a separate saved agent is optional.
Where and how to create an agent
The documented entry point is OpenAI's HTTP API, called from an application or terminal using a Platform project API key.
| Request | Result |
|---|---|
POST https://api.openai.com/v1/agents |
Saved configuration with an agent ID. |
POST https://api.openai.com/v1/agents/sessions |
Working session with its own ID and conversation. |
The simplest first request supplies agent configuration, environment: {type: openai_hosted}, and input together to the sessions endpoint. A separate saved agent is optional. To reuse one, pass agent_id when starting a session.
The beta HTTP contract requires OpenAI-Beta: agents=v1 and bearer authentication. SDKs add the beta header. This is a programmatic interface; these sources do not establish a visual agent builder in Work.