openaiDeveloperDocs

About

OpenAI Developer Docs MCP is the official public documentation server hosted by OpenAI at https://developers.openai.com/mcp. It provides read-only access to documentation published on developers.openai.com, platform.openai.com, and learn.chatgpt.com, allowing an MCP-capable agent to retrieve current source material while working.

Key Features

• Searches OpenAI developer documentation and returns relevant documentation pages.
• Fetches page content so an agent can inspect and cite the underlying documentation instead of relying only on model memory.
• Official OpenAI skills document tool names including search_openai_docs, fetch_openai_doc, list_openai_docs, list_api_endpoints, and get_openapi_spec, with some tools described as available when supported.
• Uses a hosted streamable HTTP endpoint; there is no local server package or Docker image in the official quickstart.
• Can be paired with the official OpenAI Docs Skill to encourage agents to consult Docs MCP before falling back to web search.

Use Cases

• Verify OpenAI API request schemas, required fields, endpoint details, and parameter definitions.
• Retrieve current guidance for the Responses API, Agents SDK, Realtime API, Apps SDK, Codex, and related OpenAI developer products.
• Ground coding-agent answers in official documentation and preserve source traceability.
• Reduce errors caused by outdated model knowledge when OpenAI documentation changes.

Compatibility

The official quickstart documents integrations for Codex, VS Code with GitHub Copilot Agent mode, Cursor, and Claude Code. The server itself is exposed through streamable HTTP. Exact public setup instructions are not provided in the cited guide for Claude Desktop, Windsurf, or Lovable.

Limitations

• The server is documentation-only and does not call the OpenAI API or perform actions on a user's behalf.
• It is limited to OpenAI documentation sources covered by the hosted service rather than general third-party library documentation.
• No MCP prompts, prompt templates, or separately browsable MCP resources are documented in the cited official guide.
• Client behavior still depends on whether the agent selects and invokes the documentation tools; OpenAI recommends adding persistent agent guidance when reliable use is important.

Sources

• Official guide: https://developers.openai.com/learn/docs-mcp
• Hosted endpoint: https://developers.openai.com/mcp
• OpenAI Docs Skill: https://github.com/openai/skills/blob/main/skills/.curated/openai-docs/SKILL.md
• OpenAI Knowledge Skill: https://github.com/openai/openai-agents-js/blob/main/.agents/skills/openai-knowledge/SKILL.md
• Codex MCP documentation: https://learn.chatgpt.com/docs/extend/mcp?surface=cli

MCP Setup

Prerequisites

• An MCP-capable client with remote HTTP support
• Access to the client CLI or MCP configuration file

Access requirements

• OpenAI hosts the public server at https://developers.openai.com/mcp
• No API key, OAuth flow, subscription, or environment variable is documented in the official quickstart

Companion app or plugin

No companion application is required by the official quickstart.

Client setup

Codex

Transport: streamable HTTP

codex mcp add openaiDeveloperDocs --url https://developers.openai.com/mcp

[mcp_servers.openaiDeveloperDocs]
url = "https://developers.openai.com/mcp"

• Run the command in a terminal, or add the TOML configuration to ~/.codex/config.toml.
• Run codex mcp list to verify that the server is configured.

Codex CLI and the Codex IDE extension share the same MCP configuration.

VS Code

Transport: streamable HTTP

{
  "servers": {
    "openaiDeveloperDocs": {
      "type": "http",
      "url": "https://developers.openai.com/mcp"
    }
  }
}

• Create .vscode/mcp.json in the project root and add the documented configuration.
• Open GitHub Copilot Chat, switch to Agent mode, and enable the server in the tools picker.

The official guide documents VS Code support through GitHub Copilot in Agent mode.

Cursor

Transport: streamable HTTP

{
  "mcpServers": {
    "openaiDeveloperDocs": {
      "url": "https://developers.openai.com/mcp"
    }
  }
}

• Create ~/.cursor/mcp.json on macOS or Linux and add the documented configuration.
• Restart Cursor after saving the configuration.

Cursor also has an official one-click installation link from the OpenAI Docs MCP guide.

Claude Code

Transport: streamable HTTP

claude mcp add --transport http openaiDeveloperDocs https://developers.openai.com/mcp

• Run the command from the project where Claude Code should use the server.
• Run claude mcp list to verify configuration.
• Run /mcp inside Claude Code to confirm the server is connected.

For user-wide scope, the official guide adds --scope user before the server name.

Claude Desktop

Exact public setup is not documented in the official OpenAI Docs MCP guide.

Windsurf

Exact public setup is not documented in the official OpenAI Docs MCP guide.

Lovable

Exact public setup is not documented in the official OpenAI Docs MCP guide.

Other

The hosted endpoint is documented as streamable HTTP, but exact setup for other MCP clients is not documented in the official quickstart.

Setup docs: https://developers.openai.com/learn/docs-mcp