openai-responses

The Responses API enables stateful conversations by automatically managing conversation IDs and preserving multi-turn reasoning across requests. Create conversations, send messages, and the API maintains state server-side, eliminating manual session tracking and enabling long-running agentic workflows without Chat Completions' limitations.

Built-in tools include Code Interpreter for executing code, File Search for document retrieval, Web Search for real-time information, and Image Generation. The API handles tool orchestration server-side, and you can extend capabilities with MCP server integrations for custom tools and external services.

Yes, the Responses API supports MCP integration to connect external tools and services. Define MCP tool configurations in your request, and the API manages server-side orchestration, enabling polymorphic outputs that combine reasoning, tool calls, and message streams.

Responses API migration requires adopting conversation-based workflows, conversation identifiers, and server-side tool definitions instead of client-side tool handling. The API provides automatic state management and background processing, streamlining multi-turn reasoning and long-task execution compared to Chat Completions' stateless model.

Yes, background processing is built in for long-running tasks and batch workflows. The API handles asynchronous execution while maintaining conversation state, enabling seamless integration with ML workflows and analytics pipelines without blocking request-response cycles.

Use Node.js with openai@5.19.1 or later, or edge-runtime equivalents. The API requires /v1/responses endpoints and conversation identifiers for stateful operations. Ensure your environment supports the specified SDK version for full feature compatibility including tool definitions and MCP integration.