OpenAI Agents SDK
Build agents that use tools, delegate work, and maintain context across steps.
- Languages
- Python, TypeScript
- API surface
- Application SDK
- MCP capability
- MCP client
APIs & SDKs
Define agents, tools and handoffs in your application. The SDK runs the agent loop while your application owns deployment, storage and runtime integration.
Explore the developer documentation ↗Integrations
Use function tools for application logic, specialist agents for delegation and tracing to inspect runs. The Agents SDK is distinct from the managed Agents API.
View framework integration resources ↗MCP support
The SDK guide distinguishes hosted MCP tools from runtime-managed servers. Local or private MCP servers can connect through stdio or Streamable HTTP in the application runtime.
Evidence scope: SDK-specific MCP and tracing guide. Documentation reviewed; runtime compatibility has not been tested.
Sources & review
Integration notes reviewed . Check the linked documentation for current package versions, language support and authentication requirements.
- developers.openai.com/api/docs/guides/agents/sdk ↗
- developers.openai.com/api/docs/guides/agents/integrations-observability ↗
Related frameworks
Explore frameworks in the same category or from the same maintainer.
