Prism Context Engine

The Context Operating System for Vibecoders — Eliminate “Context Pollution” by orchestrating the flow of rules, design systems, and tech stacks between the Cloud and local AI Agents.

What is Prism Context Engine?

Prism Context Engine is a complete AI-powered context governance platform that:

1. Extracts Rules from Video - Upload screen recordings, automatically extract architectural patterns using Azure OpenAI
2. Enforces Standards - Deploy rules to AI coding assistants via MCP (Model Context Protocol)
3. Enables Semantic Search - Find relevant architectural discussions using vector embeddings

The Core Metaphor

Video-to-Context Extraction

Automated Documentation: Upload a screen recording of your development workflow. Prism analyzes the video transcript using Azure OpenAI to automatically extract architectural rules, coding patterns, and best practices into actionable context that your AI coding assistant can understand.

Works with:

• Cursor ✅
• Windsurf ✅
• Claude (via MCP) ✅
• VS Code (via MCP extension)
• GitHub Copilot (coming soon)

Instead of hoping your AI follows best practices, Prism enforces them.

Quick Start

Option A: Web Dashboard (Recommended)

# 1. Sign up at the web dashboard
https://prism.jeffdev.studio/sign-up
 
# 2. Create your first project and brand profile
# 3. Upload a video or manually create rules
# 4. Export .cursorrules file and paste into your project

Option B: MCP Server (Advanced)

For real-time sync with your IDE:

# 1. Install MCP server globally
npm install -g @prism/mcp-server
 
# 2. Configure your IDE (Cursor/Windsurf/VS Code)
# Add to your MCP settings:
{
  "mcpServers": {
    "prism": {
      "command": "prism-mcp",
      "args": ["--api-key", "YOUR_API_KEY"]
    }
  }
}
 
# 3. Your AI assistant now has live access to your rules

Core Features

Architecture

┌─────────────────────────────────────────────────────────────┐
│              Video Upload (Mux)                             │
│  Screen recordings → Automatic transcription                │
└────────────────────────────┬────────────────────────────────┘
                             │ Webhook
                             ▼
┌─────────────────────────────────────────────────────────────┐
│           Azure OpenAI Processing (GPT-4o-mini)             │
│  Extract rules + Generate embeddings for search             │
└────────────────────────────┬────────────────────────────────┘
                             │
                             ▼
┌─────────────────────────────────────────────────────────────┐
│            Azure Cosmos DB (MongoDB API)                    │
│    Rules Collection + Video Transcripts + Embeddings        │
└────────────────────────────┬────────────────────────────────┘
                             │ MCP Protocol
                             ▼
┌─────────────────────────────────────────────────────────────┐
│                   Your IDE (Cursor/Windsurf)                │
│  AI follows your rules automatically during coding          │
└─────────────────────────────────────────────────────────────┘

What Makes Prism Different?

🎥 Video-First Context Capture

Record your screen while discussing architecture. Prism automatically extracts rules from your explanations.

🧠 AI-Powered Rule Extraction

GPT-4o-mini analyzes transcripts to identify:

• Architectural patterns
• Code style conventions
• Naming rules
• Testing strategies
• Security practices

🔍 Semantic Search

Vector embeddings enable finding relevant architectural discussions, not just keyword matching.

🚀 Zero-Config MCP Integration

npx prism connect and you’re done. Your AI assistant now has context.

Real-World Example

// Before Prism: Generic code
const data = await fetch('/api/users');
 
// After Prism: Enforces your standards
const { data, error } = await api.users.list();
if (error) {
  toast.error(error.message);
  return;
}

Your AI assistant learns from your videos that you:

• Use custom API wrapper with error handling
• Show errors via toast notifications
• Follow early return patterns

Built by JeffDev Studio

Prism Context Engine is part of the JeffDev  ecosystem — enterprise-grade tools for developers who want AI that respects their architecture.

Current Version: 1.0.3
License: MIT
Support: GitHub Issues