Nucleus MCP – Local-first security and identity layer for AI agents

Summary: Nucleus MCP provides a local-first security and identity layer that unifies AI agents like Cursor, Claude, and Windsurf into a single persistent identity without relying on the cloud. It features intention-aware resource locking, git-native long-term memory, and orchestration of complex tool integrations to enhance agent safety and coordination.

What it does

Nucleus MCP acts as a hypervisor with intention-aware security controls, maintains long-term memory stored in project repositories, and enables recursive mounting to manage interconnected tools such as Stripe and Postgres.

Who it's for

It is designed for developers and users running autonomous AI agents locally who need secure, persistent identity and safe orchestration of agent workflows.

Why it matters

Nucleus MCP addresses the risk of autonomous agents causing unintended damage by providing local, intention-aware security and persistent memory without cloud dependency.