How to Install Nucleus MCP in OpenAI Agents SDK

Sovereign Agent OS — Persistent Memory & Governance for AI Agents

What Nucleus MCP does

Persistent memory and governance for AI coding agents. Local-first .brain/ directory with engrams, audit log, and hypervisor file protection. Optional remote sync to relay.nucleusos.dev. Works across Cursor, Claude Code, Windsurf, and Gemini via Streamable HTTP or stdio. Features: persistent memory, governance, local-first, multi-client, remote sync. Install: pip install nucleus-mcp && nucleus-init --scan

memory

ai-agents

windsurf

Installation steps

1Install: pip install openai-agents
2Copy the snippet below into your agent code
3Replace placeholder env values with your real keys
4Pass mcp_servers=[mcp_server] to your Agent()

Python snippet

Python snippetpython

from agents import Agent
from agents.mcp import MCPServerStdio

mcp_server = MCPServerStdio(
    command="pip",
    args=["install", "nucleus-mcp", "&&", "nucleus-init", "--scan"],
)

agent = Agent(
    name="My Agent",
    model="gpt-4o",
    mcp_servers=[mcp_server],
)

Full server details GitHub Open Config Editor

Install in other clients

🤖Claude Desktop 💻Claude Code ⌨️Cursor 🏄Windsurf ♊Gemini

For maintainers

If you maintain Nucleus MCP, add this badge to your README to show it's verified on CuratedMCP:

[![CuratedMCP Verified](https://www.curatedmcp.com/api/badge/nucleus-mcp)](https://www.curatedmcp.com/marketplace/nucleus-mcp)

Related servers

Memory MCP Server

Persistent memory for AI agents using a local knowledge graph — remember facts across sessions

Free