Norn
Persistent, visible memory for AI coding agents. Norn is a local MCP server that remembers your decisions, preferences, and project context across every session and project.
README
Norn
Persistent, visible memory for AI coding agents.
Live site · Releases · Quickstart
<!-- DEMO GIF: record a short screen capture (agent remembers, then recalls in a
new session) and save it as docs/demo.gif, then replace this comment with:
-->
Your AI forgets you every session. Norn is a local MCP server that remembers your decisions, preferences, and project context across every session and project, for Claude Code, Cursor, and any MCP client. Unlike most memory tools it is fully local and fully inspectable: a dashboard lets you see exactly what it stored and forget anything you do not want.
Quickstart
Requires Node 20+. Add Norn to Claude Code in one line, no clone:
claude mcp add norn -- npx -y @samad001z/norn-server
For Cursor, Claude Desktop, or any MCP client, use the standard config:
{
"mcpServers": {
"norn": {
"command": "npx",
"args": ["-y", "@samad001z/norn-server"]
}
}
}
Restart your agent. Norn registers four tools: remember, recall, forget, list.
The first
rememberorrecalldownloads a local embedding model (all-MiniLM-L6-v2, ~25 MB) once, then runs fully offline.
<details> <summary>Or run from source</summary>
git clone https://github.com/samad001z/Norn.git
cd Norn
npm install
npm run build -w @samad001z/norn-core && npm run build -w @samad001z/norn-server
claude mcp add norn -- node "$(pwd)/server/dist/index.js"
</details>
See it work
Memory survives across sessions, and recall is semantic: it matches meaning, not keywords.
Session 1
You: Remember that we deploy to production from the main branch on Vercel.
Claude calls
remember("deploy to production from the main branch on Vercel", project: "acme")
Session 2, the next day, in a fresh context window
You: How do we ship to prod?
Claude calls
recall("how do we ship to prod")Norn returns "Deploy to production from the main branch on Vercel." even though the query shares no keywords with the stored note.
How it works
Three local pieces share one local database:
Claude Code / Cursor ──MCP (stdio)──► Norn MCP server ┐
├──► ~/.norn/norn.db
Dashboard (Next.js) ────────────────────────────┘ (SQLite + sqlite-vec)
- MCP server (
/server): exposesremember,recall,forget,listover stdio. - Store (
/core): SQLite + sqlite-vec, with embeddings from a local MiniLM model (no API key).recallblends semantic similarity with recency and trims results to a token budget;rememberdedupes near-identical notes. - Dashboard (
/web): a Next.js app to browse and manage everything.
All three default to the same file, ~/.norn/norn.db (override with NORN_DB_PATH), so
a memory written by your agent appears in the dashboard, and a memory you forget in the
dashboard is gone for the agent too.
Features
- Remembers across sessions and projects. Stop re-pasting CLAUDE.md by hand.
- See everything it knows. No black box: every memory is visible.
- Forget anything, with undo. Full control over what it keeps.
- Lives in your tools over MCP. Claude Code, Cursor, and any MCP client.
- Local-first. Your context, the embedding model, and the database all stay on your machine.
Privacy
Local by default. The store is a SQLite file on your disk, the embedding model runs on
your machine, and nothing leaves it: no account, no API key, no telemetry. Delete
~/.norn/norn.db and the memory is gone.
Manage your memories
The dashboard lets you browse by project, search, and forget any memory (with undo).
npm run dev:web
# open http://localhost:3000/app

You can also inspect the store from the command line:
npm run cli -w @samad001z/norn-core -- list
npm run cli -w @samad001z/norn-core -- recall "how do we deploy"
Roadmap
- Swappable embedding backends (Ollama, OpenAI-compatible) behind the existing
Embedderinterface. - Create and edit memories from the dashboard, not just browse and forget.
- Optional end-to-end encrypted sync, off by default.
- More editor and MCP-client integrations.
Contributing
Contributions are welcome. See CONTRIBUTING.md. The short version:
npm install
npm run build
npm test -w @samad001z/norn-core
License
MIT. See LICENSE.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。