🖱️ Mouse: Local Visualizable Task Management for Cursor

🤔 What is Mouse?

Mouse is my take on bringing structured, visual task management right into the Cursor AI IDE. It's a project ruleset I've designed to work hand-in-hand with a simple, standalone HTML file for visualizing everything. My goal here is to create a clear, manageable way for you and Cursor's AI to tackle complex tasks together. With Mouse, tasks get broken down, executed with precision, meticulously logged, and are easy to see and understand, all locally on your machine.

The core idea is simple: you need a Mouse to give Cursor precise direction. This project is my attempt to provide that "Mouse."

💡 Why Mouse? The Journey and Motivation

This project has been a bit of a journey for me. My first adventure into enhancing AI-assisted development was Rooroo, which, to my delight, became a pretty popular custom modeset for Roo code.

I genuinely love Roo Code, especially its ability to juggle multiple modes, break down big tasks into smaller, manageable ones, and prompt for key decisions. However, the cost can be a factor, which led me to create Rooroo. I wanted to build a custom modeset that was super token-efficient and unambiguous, allowing even less powerful models with smaller context windows to achieve comparable results.

But Cursor plays by different rules with its pricing model. Suddenly, I didn't have to be so conservative with context length, meaning I could afford to be more descriptive and detailed. Plus, with the ability to use powerful models like Gemini 2.5 Pro via the slow pool, I could design with the assumption that top-tier AI is available.

Another game-changer for me was Cursor's own apply diff model. It's incredibly efficient for editing large files, much more so than other tools I've used.

So, my approach shifted. I wanted to build a framework specifically for Cursor that would maximize its benefits – a system where everything is clearly logged, following a task management pattern similar to what I found effective with Rooroo.

When I looked at existing task management tools, I found many were either a hassle to set up (often requiring custom modes or MCPs), or they were overly complex with a daunting number of features. And a lot of them lacked a simple, clear way to visualize progress.

I have this philosophy of starting simple: a single file is better than multiple local files, which is better than local files with dependencies, which is better than cloud solutions. That's what led me to create this mouse setup.

Why is 🖱️ Mouse special?

I believe Mouse stands out because:

• It's refreshingly simple to get going. The visualization? Just a single HTML file (mouse_dashboard.html). You can open it directly in your browser or serve it with a basic HTTP server. No complicated dependencies or setup hoops to jump through.
• The protocol is thoughtfully crafted. I've poured a lot of thought into it, drawing inspiration from other robust prompting and task management approaches. The aim is clarity and effectiveness.
• Everything is meticulously logged. This makes it easy to pick up where you left off, recover from interruptions, and always have a clear audit trail of what the AI is doing.

🚀 Getting Started

Ready to give Mouse a try? Here's how you can get set up:

1. Clone the repo, or download it as a zip. Get the files onto your local machine in whatever way you prefer.
2. (Optional) Check out the demo results. This can give you a feel for how things look:
   1. Open mouse_dashboard.html in your browser and upload the .mouse folder (from the demo or your own initial tasks) to visualize the workflow.
   2. Alternatively, you can serve the HTML file locally. A simple way is to navigate to the directory in your terminal and run python -m http.server (if you have Python) or npx http-server (if you have Node.js/npm), then open the provided URL (usually http://localhost:8000 or similar) in your browser.
3. Copy the .cursor/rules folder into your project's root directory. This is where the core Mouse protocol and any custom project standards live, guiding the AI.
4. (Optional but Recommended) Copy mouse_dashboard.html to your project's root directory. This makes it super easy to access the visualization for your own tasks.
5. You should be all set! The AI should now be able to follow the Mouse protocol. If you find the model isn't quite following the rules as expected, you can gently remind it by referencing a specific rule file (e.g., by typing @core_protocol.mouse.mdc in the chat). This helps steer it back on course.

⚙️ How It Works (High-Level)

Want a clearer picture of how Mouse helps us collaborate? Here's a visual overview using a simple text-based diagram, followed by a breakdown:

[ User Initiates Task/Goal ]
          |
          V
+---------------------------------------------------+
| Agent (Cursor AI using Mouse Protocol)            |
|---------------------------------------------------|
| Key Guiding Principles:                           |
|   - Clarity First (Ask if unsure)                 |
|   - Log Everything Actionable                     |
|   - Protect User Data                             |
|   - Follow Task Lifecycles (Pending -> Done)      |
|   - Apply Relevant Standards                      |
+---------------------------------------------------+
          |
          V
[ 1. Task Formalization & Planning ]
          |
          |--- IF SIMPLE (S/M Effort) ---> [ Task File Created: `.mouse/tasks/TASK_ID.md` ]
          |                                    (Contains: ID, Title, Status, Description, Deliverables)
          |                                    (Direct logging of steps in `HistoryLog`)
          |
          |--- IF COMPLEX (L/X Ef