Skip to content

Quick Start

This guide walks you through setting up ResearchPad in an existing ML project in under two minutes.

1. Install ResearchPad

pip install researchpad

2. Initialize Your Project

Navigate to your ML project root and run:

researchpad init

This creates:

  • .cursor/commands/ -- Cursor IDE commands for /research, /experiment, /debug, and /explain
  • .researchpad/experiments/research/ -- Storage for research artifacts
  • .researchpad/experiments/debug/ -- Storage for debug reports
  • .researchpad/experiments/insights.md -- Accumulated experiment insights

Already have a .cursor/commands/ directory?

Use researchpad init --force to overwrite existing command files.

3. Start the UI Server

researchpad runserver

The server starts at http://localhost:8888 by default.

To bind to a different host or port:

researchpad runserver 0.0.0.0:3000

4. Open the Dashboard

Open http://localhost:8888 in your browser. You will see the ResearchPad dashboard with panels for experiments, research, debug, and insights.

The dashboard updates in real time via WebSocket -- any changes to your .researchpad/ directory are reflected immediately.

5. Use Cursor Commands

With the Cursor IDE, use these commands to interact with ResearchPad:

Command What It Does
/research Conduct background research and save a structured artifact
/experiment Run a standard experiment iteration
/experiment-bold Run an experiment with more aggressive changes
/debug Analyze a failure and generate a debug report
/explain Get a plain-language explanation of recent results

Each command reads your project context and writes structured artifacts to .researchpad/, which the dashboard picks up automatically.

What's Next

  • Configuration -- Customize storage paths and server binding
  • Features -- Explore each dashboard panel in depth
  • Commands -- Detailed reference for every command