Skip to main content
🎓Tutorial

Introduction

prAxIs OS is a portable multi-agent development framework built on praxis - the integration of theory and practice through continuous learning. It transforms AI from helpful assistant to velocity-enhancing development partner that gets better with every interaction.

What is prAxIs OS?​

prAxIs OS builds on the foundation of Agent OS by Brian Casel, addressing key limitations discovered through production use:

  • 71% fewer messages through query-first behavior (driven by MCP/RAG)
  • Architectural phase gating that enforces workflows in code
  • Universal + generated standards that work across any language
  • Specialized sub-agents for design validation, concurrency analysis, and more

Quick Start​

Install prAxIs OS in your project:

# Open your project in Cursor and say:
"Install prAxIs OS from github.com/honeyhiveai/praxis-os"

The Cursor agent will:

  1. Analyze your project (detect language, frameworks)
  2. Copy universal standards (CS fundamentals)
  3. Generate language-specific standards (tailored to your stack)
  4. Install MCP server locally
  5. Configure Cursor MCP integration
  6. Build RAG index

What Gets Installed​

your-project/
├── .cursorrules              # AI behavioral triggers (27 lines)
├── .praxis-os/
│   ├── standards/
│   │   ├── universal/        # Timeless CS fundamentals
│   │   └── development/      # Language-specific guidance
│   ├── ouroboros/           # MCP/RAG server
│   └── .cache/vector_index/  # Semantic search index
└── .cursor/
    └── mcp.json              # MCP configuration

Core Concepts​

Praxis: Theory ⇄ Practice Integration​

prAxIs OS is a praxis engine - every interaction creates learning that improves future interactions:

The Cycle:

  1. Action: AI executes task using standards
  2. Reflection: Evidence validates what worked/failed
  3. Learning: Capture pattern as new standard
  4. Refined Action: Next AI queries and applies that learning

The Result:

Session 1:  AI + 0 standards  → 70% quality
Session 50: AI + 95 standards → 95% quality

The AI doesn't get smarter. The system does.

Through accumulated praxis cycles, prAxIs OS transforms probabilistic AI behavior into deterministic quality. Each session teaches the system. Each mistake becomes impossible to repeat. Knowledge compounds across all future work.

Read more: Praxis Philosophy

MCP/RAG Architecture​

Instead of reading entire 50KB files, the MCP server uses semantic search to deliver 2-5KB targeted chunks:

  • Before: AI reads full file, 4% relevant content, 12,500 tokens
  • After: Vector search returns relevant chunks, 95% relevant, 625 tokens
  • Result: 24x better relevance per query, 71% fewer messages overall, 54% cost reduction

Universal Standards​

Timeless CS fundamentals that apply to any programming language:

  • Concurrency patterns (race conditions, deadlocks, locking strategies)
  • Testing strategies (test pyramid, test doubles, property-based testing)
  • Architecture patterns (SOLID, dependency injection, API design)
  • Failure modes (graceful degradation, circuit breakers, retry strategies)

Language-Specific Generation​

For each project, prAxIs OS generates language-specific implementations of universal standards:

  • Python: GIL, threading, asyncio, pytest patterns
  • Go: Goroutines, channels, sync primitives, table-driven tests
  • Rust: Ownership, Arc/Mutex, tokio, cargo test patterns
  • And more...

Architectural Phase Gating​

Workflows enforce progression through code, not documentation:

def can_access_phase(self, phase: int) -> bool:
    """AI cannot access Phase N+1 before completing Phase N."""
    if phase == self.current_phase or phase in self.completed_phases:
        return True
    return False  # Structurally impossible to skip

Next Steps​