AI-Assisted Development

At paiOS, we embrace AI-assisted development. We use tools like Cursor, Copilot, and TaskMasterAI to stay productive and focus on high-value work.

Recommended MCP Servers

For the best AI-assisted development experience, we recommend using these MCP (Model Context Protocol) servers:

• Context7 - Provides up-to-date documentation and code examples for libraries and frameworks. Essential for staying current with best practices and API changes.
• TaskMasterAI - Manages task breakdown, progress tracking, and provides structured context for AI agents working on complex features.

These MCP servers integrate directly with AI coding assistants (like Cursor) and provide rich, structured context without requiring manual file management.

The Hybrid Workflow

To maximize efficiency with AI, we follow a three-layered workflow:

1. GitHub Issues (Status): The single source of truth for task status. Every feature or bug fix starts here.
2. Starlight Docs (Specs): The source of truth for architecture and technical specifications. Refer to these guides to provide context to your AI assistant.
3. MCP Servers (Primary Context): Use MCP servers (Context7, TaskMasterAI) to provide structured, up-to-date context directly to your AI assistant. This is the preferred method.
4. Local Context (Optional): For very specific, temporary execution details, you can still use git-ignored context files, though MCP servers often make this unnecessary.

Using TaskMasterAI

For larger or more complex tasks, we strongly recommend using TaskMasterAI. It helps organize the work into manageable steps and provides a structured context for AI agents.

How to use it:

• Initialize: If you are working on a new feature, TaskMasterAI can help you break down a PRD into tasks.
• Context Management: TaskMasterAI generates context files that can be directly referenced by AI coding assistants.
• Progress Tracking: Use TaskMasterAI to log progress, findings, and decisions as you implement each subtask.

Refer to the TaskMasterAI Documentation for more details on the CLI and MCP tools.

AI Coding Assistant Guidelines

When working with AI assistants (like Cursor), follow these best practices to ensure high-quality contributions:

1. Provide High-Quality Context

AI tools are only as good as the context you provide. Before starting a task:

• Use MCP Servers: Configure Context7 and TaskMasterAI MCP servers in your AI assistant (e.g., Cursor) to automatically provide structured context.
• Reference the relevant Starlight documentation (Architecture, Guides).
• Use TaskMasterAI to maintain a clear list of subtasks and provide task-specific context via MCP.
• Optional: For very specific, temporary execution details, you can still use local spec files (e.g., .cursor/active_task.md), though MCP servers often make this unnecessary.

2. Review and Refine

NEVER blindly accept AI-generated code. Always:

• Review the code for adherence to our Development Standards (KISS, SOLID).
• Check for “hallucinated” APIs or unnecessary dependencies.
• Ensure the code is readable and well-documented.

3. Maintain the “Human in the Loop”

The AI is your assistant, not the primary developer. You are responsible for the architectural integrity and long-term maintainability of the code.

Local Execution Context (Optional)

With MCP servers (Context7, TaskMasterAI) configured, you typically don’t need local context files. The MCP servers provide structured, up-to-date context directly to your AI assistant.

However, for very specific, temporary execution details that aren’t captured in GitHub Issues or TaskMasterAI, you can still create a temporary file:

// .cursor/active_task.md (git-ignored)
# Task: Implement VAD (Voice Activity Detection)
- Issue: #123
- Architecture: [Link to docs.aurintex.com/architecture/hal]
- Goal: Capture audio and trigger transcript only when speech is detected.
- Constraints: Must use `webrtc-vad` crate.

Note: This is rarely necessary when using MCP servers. Most context should come from:

• GitHub Issues (via TaskMasterAI MCP)
• Starlight documentation (via Context7 MCP for library docs)
• TaskMasterAI task context (via TaskMasterAI MCP)

Once the task is done and merged, delete any temporary files you created.

---

By following this workflow, you can leverage the power of AI while maintaining the high standards of the paiOS project. 🚀