This tutorial will guide you through installing and using HoloViz MCP with Claude Code (the command-line interface). By the end, you'll have HoloViz MCP running and be able to ask Claude questions about Panel components directly from your terminal!
!!! tip "What you'll learn" - How to install HoloViz MCP - How to configure it with Claude Code - How to use it to get help building Panel applications - How to verify everything is working correctly - How to build your first Panel dashboard
!!! note "Prerequisites" Before you begin, ensure you have:
- **Python 3.11 or newer** installed on your system
- **[uv](https://docs.astral.sh/uv/)** package installer
- **Claude Code CLI** installed ([Installation guide](https://claude.ai/download))
- **panel** installed and runnable in your terminal
Open your terminal and install HoloViz MCP as a uv tool:
uv tool install "holoviz-mcp[pydata]"This command installs HoloViz MCP globally, making it available for Claude Code to reference.
!!! tip "What's happening?" The uv tool manager creates an isolated environment for HoloViz MCP and installs all necessary dependencies.
The extra `pydata` dependencies are added to install a wide range of python data related packages. We will assume these are installed throughout this guide. You can replace them with your favorite dependencies for your own work.
Install Chromium to enable the holoviz-mcp server to take screenshots:
holoviz-mcp install chromium📦 This downloads 300MB as it downloads the Chromium and FFMPEG engines.
HoloViz MCP needs to index the HoloViz documentation to provide intelligent answers. Run:
holoviz-mcp update index⏱️ This will take 5-10 minutes as it downloads and indexes documentation from Panel, hvPlot, and other HoloViz libraries.
Now let's configure Claude Code to use the HoloViz MCP server:
claude mcp add holoviz --transport stdio --scope user -- holoviz-mcpThis will update your ~/.claude.json file with the HoloViz MCP server configuration.
HoloViz MCP includes specialized agents for Claude Code that help with planning and implementing HoloViz applications.
Navigate to your project directory and run:
holoviz-mcp install claudeThis creates a .claude/agents/ directory with:
holoviz-data-explorer.md- Agent for quick data exploration and data visualizationholoviz-app-architect.md- Agent for architecting production Panel data applications
!!! tip "Install User-Level Agents"
To make agents available across all your projects:
```bash
holoviz-mcp install claude --scope user
```
This installs agents to `~/.claude/agents/`.
Let's verify that HoloViz MCP is working correctly!
Start Claude Code and run the /mcp command to verify the status of the HoloViz MCP server:
/mcpYou should see holoviz listed as an available MCP server.
Press Escape to return to the prompt.
Open a chat with Claude Code and try these questions:
Component Discovery:
What Panel components are available for user input?
You should see it using the panel_search_components tool:
Component Details:
What parameters does the Panel Button component accept?
If Claude provides detailed, accurate answers with specific Panel component information, congratulations! HoloViz MCP is working correctly! 🎉
Now that everything is set up, let's build a simple dashboard.
Ask Claude:
Create a Panel dashboard that displays a slider and shows the square of the slider's value. Save it to app.py
Claude will provide code using HoloViz MCP's knowledge of Panel components!
Save the code to app.py and run it:
run it!!! tip "Panel is not installed" If the command errors with "Panel is not installed" you will have to create a virtual environment and install panel
Your dashboard will open in your default web browser!
HoloViz MCP includes a powerful display tool that can render visualizations directly. Ask Claude:
Use the holoviz_display tool to show me a simple hvplot visualization of random data.Claude will use the display tool to generate and display the visualization. See the Display System tutorial for more details.
Now that you have HoloViz MCP running with Claude Code, explore more:
- Display System: Learn about the display server for visualizations
- Stock Analysis: Create a real-world stock analysis report
- Weather Dashboard: Create an interactive weather visualization
Problem: uv: command not found
Solution: Install uv by following the uv installation guide
Problem: claude: command not found
Solution: Install Claude Code by following the installation guide
Problem: Installation takes too long
Solution: This is normal! The first installation downloads many dependencies. Subsequent updates are much faster.
Problem: /mcp command doesn't show holoviz server
Solution:
- Check that the configuration command completed successfully
- Verify
~/.claude.jsoncontains the holoviz server configuration - Try running the server directly:
holoviz-mcp - Run
claude mcp listto see configured servers
Problem: Claude doesn't recognize Panel components
Solution:
- Check that the documentation index completed (Step 3)
- Verify your configuration is correct:
claude mcp list - Try running the server directly in terminal:
holoviz-mcp - Check Claude Code logs for errors
Problem: MCP server won't connect
Solution:
- Verify Python 3.11+ is installed:
python --version - Check uv installation:
uv --version - Try running the server directly:
holoviz-mcp - Check the Claude Code logs for errors
For more help, see the Troubleshooting Guide or join the HoloViz Discord.
In this tutorial, you:
- ✅ Installed HoloViz MCP using uv
- ✅ Created the documentation index
- ✅ Installed Chromium
- ✅ Configured Claude Code
- ✅ Verified the installation
- ✅ Built your first Panel dashboard
- ✅ Learned about the display tool
- ✅ Learned how to work on projects with Claude Code
You're now ready to use HoloViz MCP with Claude Code to accelerate your Panel development from the command line! Happy coding! 🚀