# NotebookLM MCP — CLI + 35 Tools for Agents > NotebookLM MCP (notebooklm-mcp-cli) gives you a CLI (`nlm`) and an MCP server (`notebooklm-mcp`) so assistants can operate Google NotebookLM notebooks. ## Install Merge the JSON below into your `.mcp.json`: ## Quick Use 1. Install / set up: ```bash pip install notebooklm-mcp-cli ``` 2. Start / smoke test: ```bash nlm login && nlm setup add claude-code ``` 3. Verify: - Run `nlm doctor` and confirm installation/auth/config checks pass for your client. ## Intro NotebookLM MCP (notebooklm-mcp-cli) gives you a CLI (`nlm`) and an MCP server (`notebooklm-mcp`) so assistants can operate Google NotebookLM notebooks. - **Best for:** research/coding workflows that want NotebookLM notebooks accessible from Claude Code, Cursor, Gemini CLI, and similar tools - **Works with:** Python tooling (pip/pipx/uv), a Chromium-based browser for login, MCP-compatible clients - **Setup time:** 12 minutes ## Practical Notes - Exposes 35 MCP tools (disable it when not needed to preserve context window) - One install yields both `nlm` (CLI) and `notebooklm-mcp` (MCP server) - GitHub stars/forks (verified): see Source & Thanks If you want NotebookLM inside your agent workflow, this project gives you two practical layers: - **CLI layer (`nlm`)** for scripting: list/create notebooks, add sources, share, export, etc. - **MCP layer (`notebooklm-mcp`)** for natural-language tool calls from your assistant. A reliable setup path: 1. Install via pip/pipx/uv. 2. Authenticate with `nlm login` (auto mode launches a dedicated browser profile and persists cookies). 3. Use `nlm setup add ` to configure your AI tool without hand-editing JSON. Because it exposes many tools, keep it *off* when you’re not actively using NotebookLM — it’s easy to burn context budget accidentally. ### FAQ **Q: How do I authenticate?** A: Run `nlm login` (auto mode launches a browser and extracts cookies). **Q: Which AI tools are supported?** A: Use `nlm setup add ` for supported clients, or generate JSON with `nlm setup add json`. **Q: How do I troubleshoot?** A: Run `nlm doctor` to diagnose installation, auth, and client configuration issues. ## Source & Thanks > Source: https://github.com/jacob-bd/notebooklm-mcp-cli > License: MIT > GitHub stars: 4,309 · forks: 688 --- ## 快速使用 1. 安装 / 设置: ```bash pip install notebooklm-mcp-cli ``` 2. 启动 / 冒烟测试: ```bash nlm login && nlm setup add claude-code ``` 3. 验证: - 运行 `nlm doctor`,确认安装/认证/配置检查对你的客户端均通过。 ## 简介 NotebookLM MCP(notebooklm-mcp-cli)把 AI 工具接到 Google NotebookLM:一次安装同时获得 CLI(`nlm`)与 MCP server(`notebooklm-mcp`),并提供 35 个可调用工具管理 notebook 与资料来源。 - **适合谁:** 需要在 Claude Code/Cursor/Gemini CLI 等工具里直接操作 NotebookLM 的研究/开发工作流 - **可搭配:** Python 工具链(pip/pipx/uv)、Chromium 系浏览器登录、MCP 兼容客户端 - **准备时间:** 12 分钟 ## 实战建议 - 提供 35 个 MCP 工具(不用时建议关闭以节省上下文窗口) - 一次安装同时获得 `nlm`(CLI)与 `notebooklm-mcp`(MCP server) - GitHub stars / forks(已核验):见「来源与感谢」 想把 NotebookLM 融进 Agent 工作流时,这个项目提供两层能力: - **CLI(`nlm`)** 用于脚本化:列出/创建 notebook、添加来源、分享与导出等。 - **MCP(`notebooklm-mcp`)** 让助手用自然语言调用工具完成同样的操作。 更稳的配置路径: 1. 用 pip/pipx/uv 安装。 2. `nlm login` 进行认证(自动模式会启动独立浏览器 profile 并持久化 cookies)。 3. `nlm setup add ` 自动写入对应客户端配置,避免手改 JSON。 由于工具数较多,不使用 NotebookLM 时建议关闭它,避免无意间占用上下文预算。 ### FAQ **怎么登录认证?** 答:运行 `nlm login`(自动模式会打开浏览器并提取 cookies)。 **支持哪些 AI 工具?** 答:对已支持的客户端用 `nlm setup add `;其它工具可用 `nlm setup add json` 生成配置。 **遇到问题怎么排查?** 答:运行 `nlm doctor` 诊断安装/认证/客户端配置。 ## 来源与感谢 > Source: https://github.com/jacob-bd/notebooklm-mcp-cli > License: MIT > GitHub stars: 4,309 · forks: 688 --- Source: https://tokrepo.com/en/workflows/notebooklm-mcp-cli-35-tools-for-agents Author: MCP Hub