Hosting

Local MCP server

Transport

stdio

Endpoint

Local process; no hosted endpoint documented

Authentication

No authentication is configured. The maintainers say to run the server locally because tools can read local data and the server has no access control.

Requirements

• Docker, or uv and the uvx executable.
• For local datasets in Docker, mount only the required files or directories into the container.
• An MCP host that supports tool calling.

Setup steps

1. Install Docker using its official instructions, then configure the MCP client to run `docker` with args [`run`, `-i`, `--rm`, `mcp/vizro`] over stdio.
2. Alternatively, install uv and configure the client to run the installed `uvx` executable with args [`vizro-mcp`], as shown in the official setup guide.
3. Reload the MCP server and inspect tools/list. The official API reference documents nine tools, including `create_starter_dashboard` and `get_sample_data_info`.
4. For local data with Docker, add a bind mount for only the specific allowed data directory or file, using the documented `--mount type=bind,src=<absolute-path>,dst=<absolute-path>` form.
5. For a safe initial check, call `get_sample_data_info` on a documented sample dataset. Avoid using private data with preview actions unless you have reviewed where the data will be sent.

The official API reference enumerates nine tools. This is documentation-derived; the server's live tools/list was not observed.

• create_dashboard

  Create an exploratory dashboard based on one local or remote dataset.

  Parameters: file_path_or_url, context (optional)
  01
• create_starter_dashboard

  Provide a starter dashboard example.

  02
• create_vizro_chart

  Create a Vizro-styled chart from a local or remote dataset.

  Parameters: file_path_or_url, context (optional)
  03
• get_model_json_schema

  Return the JSON schema for a Vizro model.

  Parameters: model_name
  04
• get_sample_data_info

  Return metadata for a documented sample dataset.

  Parameters: data_name
  05
• get_vizro_chart_or_dashboard_plan

  Return instructions and a plan for creating a Vizro chart or dashboard.

  Parameters: user_plan, user_host, advanced_mode (optional)
  06
• load_and_analyze_data

  Load and analyze data from an absolute local path or URL.

  Parameters: path_or_url
  07
• validate_chart_code

  Validate chart code and optionally open a PyCafe preview.

  Parameters: chart_config, data_info, auto_open (optional)
  08
• validate_dashboard_config

  Validate a dashboard configuration and optionally open a PyCafe preview.

  Parameters: dashboard_config, data_infos, custom_charts, auto_open (optional)
  09

This server’s available tools can vary by account, permissions, or configuration. The list shown is not its full runtime catalog.

• Connection and a representative function call were not independently tested for this catalog record.
• The maintainers state that the server currently has no authentication and should be run locally, not hosted remotely.
• The `validate_dashboard_config` and related preview behavior can open a PyCafe link; do not send private datasets to a remote preview service without reviewing the data flow.
• For Docker, local files are not visible unless explicitly mounted. Keep mounts limited to data the model is authorized to read.

• official mcp setup documentation

  Documents the local stdio setup using Docker image `mcp/vizro` or uvx package `vizro-mcp`, and bind-mount syntax for local files.

• official mcp tool reference

  Enumerates nine MCP tools, describes their parameters, warns that the unauthenticated server should be run locally, and notes the PyCafe preview behavior.

• official repository

  The McKinsey/Vizro repository identifies Vizro-MCP as part of the official Vizro project and links to its documentation; the repository showed about 3.8k stars when checked.

• verified publisher organization avatar

  McKinsey organization avatar downloaded, decoded as a 300px image, and visually reviewed on light and dark contact sheets. Selected fallback tier B: developer organization identity; the repository README's Vizro image is a wide banner, not a suitable standalone card icon.

• discovery directory

  The October 17 research shortlist included the McKinsey/Vizro repository as an MCP candidate; the directory was used only for discovery, not as proof of setup or capabilities.

• official or maintainer tool reference

  Official API reference enumerates nine tools; catalog count and rows are exact.

Documentation checked 2026-10-28T00:00:00Z. Connection has not been runtime tested.