# jupyter-mcp-server — MCP Server for Jupyter

> jupyter-mcp-server is an MCP server for Jupyter: agents can list files, run cells, and cite notebooks with verified 1,090★ and OpenTelemetry hooks.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

```bash
pip install uv
pip install jupyterlab==4.4.1 jupyter-collaboration==4.0.2 jupyter-mcp-tools>=0.1.4 ipykernel pycrdt
jupyter lab --port 8888 --IdentityProvider.token MY_TOKEN
# MCP client: command=uvx args=["jupyter-mcp-server@latest"] env JUPYTER_URL=http://localhost:8888 JUPYTER_TOKEN=MY_TOKEN
```

## Intro

jupyter-mcp-server is an MCP server for Jupyter: agents can list files, run cells, and cite notebooks with verified 1,090★ and OpenTelemetry hooks.

**Best for:** Data/science teams who want agents to execute notebooks with traceable tool calls

**Works with:** Any MCP client + JupyterLab/Jupyter server (uvx or Docker modes)

**Setup time:** 8-15 minutes

### Key facts (verified)

- GitHub: 1090 stars · 162 forks · pushed 2026-05-13.
- License: BSD-3-Clause · owner avatar + repo URL verified via GitHub API.
- README-verified entrypoint: `uvx jupyter-mcp-server@latest`.

## Main

- Start with the uvx quick-start and a single notebook: validate `list_files` + a simple cell execution before adding more tools.

- Use the built-in observability hooks (README mentions OpenTelemetry) when you need reproducible traces for notebook runs and tool calls.

- Keep tokens under control by asking the agent to cite specific cells (README documents prompts/tools) instead of pasting full notebooks into chat.

### Source-backed notes

- README documents uvx and Docker client configuration for MCP, and notes a breaking change requiring `MCP_TOKEN` in v1.0.0+.
- README lists notebook tools (file/kernels/cell execution) and mentions multimodal outputs (plots/images/text).
- README highlights built-in observability via hooks + OpenTelemetry integration for tracing tool calls and kernel execution.

### FAQ

- **Do I need Docker to use it?**: No — README provides a uvx quick start; Docker is recommended for production isolation.
- **What do I need to configure first?**: A running Jupyter server + your MCP client config (JUPYTER_URL/TOKEN); README also notes `MCP_TOKEN` for newer versions.
- **Does it support tracing?**: Yes — README mentions OpenTelemetry hooks for notebook/tool execution traces.

## Source & Thanks

> Source: https://github.com/datalayer/jupyter-mcp-server
> License: BSD-3-Clause
> GitHub stars: 1090 · forks: 162

---

<!-- ZH -->

## Quick Use

```bash
pip install uv
pip install jupyterlab==4.4.1 jupyter-collaboration==4.0.2 jupyter-mcp-tools>=0.1.4 ipykernel pycrdt
jupyter lab --port 8888 --IdentityProvider.token MY_TOKEN
# MCP client: command=uvx args=["jupyter-mcp-server@latest"] env JUPYTER_URL=http://localhost:8888 JUPYTER_TOKEN=MY_TOKEN
```

## Intro

jupyter-mcp-server 是 Jupyter 的 MCP server：让 agent 可列文件、运行单元格并引用笔记本，已验证 1,090★，并内置 OpenTelemetry hook。

**Best for:** 希望让 agent 可控执行 notebook 且可追踪调用链的数据/科研团队

**Works with:** 任意 MCP 客户端 + JupyterLab/Jupyter Server（uvx 或 Docker 模式）

**Setup time:** 8-15 minutes

### Key facts (verified)

- GitHub：1090 stars · 162 forks；最近更新 2026-05-13。
- 许可证：BSD-3-Clause；作者头像与仓库链接均已通过 GitHub API 复核。
- README 中核对过的入口命令：`uvx jupyter-mcp-server@latest`。

## Main

- 先按 README 的 uvx 快速上手跑通一个 notebook：验证 `list_files` 与最小单元格执行，再逐步开放更多工具。

- 需要可复现与可审计时，启用内置可观测 hook（README 提到 OpenTelemetry），为 notebook 执行链路留痕。

- 让 agent 用“引用具体 cell”的方式工作（README 有 prompts/tools），避免把整本 notebook 粘进对话导致 token 爆炸。

### Source-backed notes

- README 给出 uvx 与 Docker 的 MCP 客户端配置，并提示 v1.0.0+ 需要设置 `MCP_TOKEN` 的 breaking change。
- README 列出文件/内核/单元格执行等工具能力，并说明支持图像/图表/文本等多模态输出。
- README 强调可观测性：通过 hook + OpenTelemetry 追踪工具调用与 kernel 执行。

### FAQ

- **必须用 Docker 吗？**：不必。README 提供 uvx 快速上手；Docker 更适合生产隔离。
- **最先要配置什么？**：先启动 Jupyter Server，再在 MCP 客户端里填 JUPYTER_URL/TOKEN；README 也提示新版本需要 `MCP_TOKEN`。
- **支持可观测/追踪吗？**：支持。README 提到通过 OpenTelemetry hook 记录 notebook 与工具执行链路。

## Source & Thanks

> Source: https://github.com/datalayer/jupyter-mcp-server
> License: BSD-3-Clause
> GitHub stars: 1090 · forks: 162


---
Source: https://tokrepo.com/en/workflows/jupyter-mcp-server-mcp-server-for-jupyter
Author: MCP Hub