OpenAI Responses

About

This Model Context Protocol (MCP) server leverages the OpenAI Responses API as its core inference engine, designed for seamless integration with MCP clients like Claude Code and OpenAI Codex. It distinguishes itself by always permitting `web_search`, allowing the underlying model to autonomously decide when and how to perform searches. The server provides structured outputs, including the main response body, a flag indicating `used_search`, and citations that specify information sources (URLs or source IDs). With a flexible configuration system and robust debugging options, it acts as a powerful backend for AI agents requiring advanced reasoning and up-to-date information retrieval.

Key Features

• Utilizes the official OpenAI JS SDK for Responses API compliance.
• Grants models autonomous control over web search functionality (`web_search` always permitted).
• Implements MCP stdio for `initialize`, `tools/list`, and `tools/call` operations.
• Supports external System Policies via configuration files.
• 0 GitHub stars
• Provides structured outputs including response body, `used_search` flag, and `citations` with source URLs.

Use Cases

• Integrating with MCP clients like Claude Code or OpenAI Codex for AI agent interaction.
• Providing structured, cited responses for tasks requiring external data verification.
• Empowering AI agents to perform autonomous web searches for problem-solving and information gathering.