N2N Memory
Ein lokaler MCP-Memory-Server, der dauerhaftes Projektwissen im jeweiligen Repository speichert und Kontextvermischung verhindert.
n2n-memory turns project context into a local asset instead of a chat-session side effect. Each repository can keep durable AI memory in .mcp/memory.json and active task context in .mcp/context.json, so agents can recover architectural decisions and project-specific pitfalls without mixing one repository with another.
The product emphasizes project isolation, Git reviewability, and explicit governance. Teams can commit durable memory when sharing is intentional, keep active context local, and use MCP tools for graph reads, writes, search, deletion, and export. That makes the tool most useful for developers who move between many codebases and need AI assistants to remember the right things in the right place.
Project Memory Model
Reviewable Storage
Keep durable memory in files that teams can inspect and optionally share through Git.
Knowledge Graph
Record entities, observations, and relations that AI agents can read, search, and update.
MCP Tooling
Expose focused tools for reads, writes, search, maintenance, and Markdown export.
Project Isolation
Store memory under each repository so context does not bleed between unrelated projects.
MCP client setup
{
"mcpServers": {
"n2n-memory": {
"command": "npx",
"args": ["-y", "n2n-memory"]
}
}
}
Documentation