Vibe Coding with Kedro-MCP¶

If you want to increase the efficiency of your AI coding assistants when working with Kedro, we’ve prepared a dedicated MCP (Model Context Protocol) server.

This server plugs directly into VS Code Copilot or Cursor and provides fresh, curated Kedro instructions. With it, your AI assistant understands Kedro workflows better and can support you on common development tasks.

Quick Install¶

To enable Kedro MCP tools in your editor, click one of the links below. Your editor (VS Code / Cursor) will open automatically and you need to press the Install button.

Once installed, your AI assistant gains access to Kedro-specific MCP tools.

Helpful references¶

Universal configuration (JSON)¶

You can also reuse this configuration in any editor or MCP-compatible assistant (e.g. Claude, Cursor, Copilot):

{
  "command": "uvx",
  "args": ["kedro-mcp@latest"],
  "env": {
    "FASTMCP_LOG_LEVEL": "ERROR"
  },
  "disabled": false,
  "autoApprove": []
}

Usage¶

After kedro-mcp installation, open Copilot Chat Agent Mode (in VS Code) or the Chat panel (in Cursor). Type / to see available MCP prompts.

Convert a Jupyter Notebook into a Kedro Project¶

/mcp.Kedro.convert_notebook

When you run this prompt, your assistant will explicitly call the MCP server and follow the provided guidance.

What happens step by step:

The assistant will analyse your Jupyter Notebook (make sure to provide the notebook content as context, or adjust the prompt to reference its filename).
It will create a conversion plan (Statement of Work) and save it into your working directory as a .md file.
You’ll be able to review the plan in chat and modify it if needed.
Once you approve the plan, the assistant will:
- Check that a virtual environment (venv) is active. If not, it will propose creating one.
- Ensure the latest version of Kedro is installed; install it if missing.
- Scaffold a Kedro project with kedro new.
- Define pipelines with kedro pipeline create.
- Populate parameters.yml and catalog.yml according to the plan.

Tips:

You can edit the prompt itself if you want to specify a particular virtual environment (e.g. conda, uv, or venv).
If any errors occur during conversion, ask the assistant in chat how to resolve them — it will guide you through fixes.
The assistant will only proceed with implementation after you’ve signed off on the plan.

Kedro Project Migration¶

/mcp.Kedro.project_migration

When you run this prompt, the assistant will call the MCP server to guide you through migrating an existing project to the latest Kedro version.

Step 1 — Plan: It analyses your project and suggests a migration plan (e.g. 0.19 → 1.0). Waits for approval.
Step 2 — Build: After approval, it ensures a virtual environment (venv) is active. If not, it creates one. The assistant will install Kedro if missing and then follows the plan.

Use this prompt to get up-to-date migration tips and avoid outdated patterns. If errors appear, ask the assistant in chat how to resolve them.

General Kedro questions¶

/mcp.Kedro.general_usage

In this case, you can type any Kedro-related request after the header line. The MCP server will then provide the assistant with the latest Kedro guidance and conventions, ensuring answers are based on up-to-date best practices.

Tips:

Keep the header line as is (so the MCP server is explicitly called).
Replace the placeholder with your actual request (e.g., “generate me a sample Kedro project for a sales forecasting pipeline”).
The assistant will then use the fresh Kedro instructions (e.g., conventions, workflow, catalog setup, parameters, QA rules) to guide you.

With Kedro MCP, Copilot and Cursor become much smarter about Kedro — so you can focus on building pipelines, not fixing AI mistakes.