
Vizro MCP Server: Setup for Claude, Cursor and Other Apps
Lets Claude, Cursor and other assistants build Vizro charts and dashboards from your data files and check that the configuration is valid.
This is the Vizro team's own server, part of the McKinsey Vizro project. It runs on your computer, either through Docker or with uv, and needs no account or API key.
It works with a chat assistant. The assistant writes the dashboard, and the server checks it and hands back the Python code. For remote data it can also open a PyCafe preview link in your browser.
Good for
- Analysts who want a Python dashboard from a CSV without writing it by hand
- Data scientists who already use Vizro and want an assistant to draft configs
- Teams who want a quick dashboard preview they can share by link
Works in
12 should work
Tap an app to see how to set it up and any known problems.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Client does not support this transport.
Client does not support this transport.
Install vizro MCP
Claude Code
- 1
Copy this prompt.
Add the vizro MCP server to Claude Code. Run this command: claude mcp add vizro -- uvx vizro-mcp Then check that it loaded and tell me if I need to sign in or restart.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires uv (Python).
Read-only mode: The server only reads data, so there is no setting to change.
Both methods are documented by Vizro. Use uvx if you have uv installed, or Docker if you prefer a container. With Docker, mount the folders that hold your data files, as the README shows.
Built from the package data we hold. Not yet tested by us. Source.
Tools
Gives the assistant a plan for building a chart or dashboard
Returns step-by-step instructions for creating a Vizro chart or dashboard. It is meant to be called first, and again in advanced mode only when custom CSS, components or actions are needed.
Try: “Build me a Vizro dashboard from my sales CSV.”
Read ·
get_vizro_chart_or_dashboard_planOffers sample data to build with
Returns information about built-in sample data for when you have no data of your own.
Try: “I have no data. Use a sample dataset.”
Read ·
get_sample_data_infoReads a data file and describes its columns
Loads a local file or a web address and returns its columns, types and a summary. It needs an absolute file path or a URL.
Try: “Look at /Users/me/sales.csv and tell me what columns it has.”
Files ·
load_and_analyze_dataLooks up the settings a Vizro component accepts
Returns the JSON schema for a given Vizro model, matching the server's Vizro version, so the assistant writes valid configuration.
Try: “What options does a Vizro Graph accept?”
Read ·
get_model_json_schemaChecks a dashboard config and returns the code
Validates a Vizro dashboard configuration and, if valid, returns the Python code. When all data is remote it also returns a PyCafe link and can open it in your browser.
Try: “Check this dashboard config and open a preview.”
Network ·
validate_dashboard_configChecks a chart and returns its code
Validates a chart plan and returns the Python code. When the data is remote it builds a PyCafe link and can open it in your browser.
Try: “Validate this bar chart and show it to me.”
Network ·
validate_chart_code
Read from the source code
Permissions
low risk · 2 things to watch
Low risk
It reads the data files you point it at and can open a link in your browser. The README says it only reads data and never writes, deletes or modifies data on your machine.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
It reads files you point it at
With Docker you choose which folders to mount, and with uv it can read any absolute path the assistant gives it. Mount only the data you need. github.com
It can open preview links in your browser
For remote data it builds a PyCafe link and may open it. Turn off auto-open by telling the assistant not to open it. github.com
Security advisories
No published security advisories as of Nov 1, 2026.
Technical details
| Talks to | Web addresses you give it for data. py.cafe (preview links) |
|---|---|
| Sign-in and keys | No account or key. |
Questions
Is Vizro MCP official?
Yes. It ships in the mckinsey/vizro repository, from the team that makes Vizro.
Do I need an API key?
No. You need either Docker or uv installed, and nothing else.
How do I use my own data files?
With Docker, mount the folder or file into the container using the extended config in the README. With uv, give the assistant the absolute path.
Should I use Docker or uv?
Both are documented, and the install box offers both. Use uvx if you already have uv, or Docker if you prefer a container.
Can it write files?
The README says the server only reads data and never writes, deletes or modifies data on your machine. The assistant itself may still write an app.py if your client allows it.
Which model works best?
The README suggests a high-performing model such as claude-4-sonnet and warns that an auto setting can be inconsistent.
Information
updated Nov 2, 2026
- Publisher
- Official
- Package
vizro-mcp- Registry
- PyPI
- Transport
- stdio
- Auth
- None
- License
- Apache-2.0
- Last commit
- Oct 30, 2026
- Open issues
- 2
- Listed in
- Docker MCP Catalog
Sources
Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.
Show all 2Hide
- README.md · github.com
- https://pypi.org/project/vizro-mcp/ · pypi.org