AI Memory OS 🧠⚡

Supercharge your VS Code with a local, intelligent memory system designed for both humans and AI Agents.

AI Memory OS is a high-performance workspace intelligence extension that bridges the gap between Structural AST Analysis and Semantic Vector Retrieval. It gives you (and your AI assistants) persistent, token-efficient context about your entire codebase with real-time synchronization.

Perfect for: VS Code users, AI agent integration, developers working with large codebases, Claude Code integration, workspace intelligence, semantic code search.

---

🚀 Key Features

1. Hybrid Intelligence Stack

Combines web-tree-sitter (AST) for deterministic dependency mapping with LanceDB (Vector RAG) for semantic search. It doesn't just find words; it understands code relationships.

2. Continuous Intelligence (Real-time Sync)

Our ContinuousDiffEngine monitors your filesystem and updates the memory index in the background within 2 seconds of any change. No manual re-indexing required.

3. Agent-First Design (Headless CLI)

Built specifically for agents like Antigravity and Claude Code. Exposes a dedicated CLI bridge so AI agents can "plug in" to your project's local memory to answer questions faster and cheaper.

4. Visual Dependency Graph 🗺️

Generate high-fidelity Mermaid.js diagrams of your project's architecture directly inside VS Code. See exactly how files are connected and how a change in one place impacts the rest.

5. Token-Efficient "Smart Answers" 🧠

Our built-in Cache Manager prunes and optimizes code context before sending it to the LLM, ensuring you stay well within Free Tier token limits (like the 150k limit) while getting Pro-level answers.

6. MCP Server Integration

Seamlessly integrate with Claude Code through the Model Context Protocol (MCP). Add to your claude.json:

{
  "mcpServers": {
    "ai-memory": {
      "command": "node",
      "args": ["${workspaceFolder}/out/mcp_server.js"],
      "env": {
        "WORKSPACE": "${workspaceFolder}"
      }
    }
  }
}

Available MCP Tools:

• ai_memory_status - Check indexing status
• ai_memory_project_map - Get high-level directory map
• ai_memory_search - Semantic vector search (CRITICAL for token savings)
• ai_memory_impact - Dependency graph analysis
• ai_memory_answer - Get grounded answers about your codebase

---

📦 Tech Stack

Component	Technology
Parsing	web-tree-sitter (WASM)
Vector Database	@lancedb/lancedb + apache-arrow
Languages Supported	TypeScript, JavaScript, Python
Interface	VS Code Webview + Mermaid.js
AI Integration	Gemini API, HuggingFace Transformers.js
Protocol	Model Context Protocol (MCP)

---

🛠️ Commands

Command	Description
AI Memory: Start Service	Initializes local indexing
AI Memory: Ask a Question	Synthesize answers from your project memory
AI Memory: Reindex Workspace	Force full workspace reindexing
AI Memory: Show Index Status	View current indexing progress
AI Memory: Set Gemini API Key	Configure AI response generation
AI Memory: Analyze Impact	See which files are affected by changes
AI Memory: Show Project Map	View high-level project structure
AI Memory: Show Dependency Graph	Visualize project architecture

---

⚙️ Configuration

Customize AI Memory OS via VS Code settings:

{
  "aiMemory.autoIndexOnStartup": true,
  "aiMemory.geminiModel": "gemini-2.5-flash",
  "aiMemory.embeddingModel": "Xenova/all-MiniLM-L6-v2",
  "aiMemory.embeddingDimensions": 384,
  "aiMemory.maxIndexedFileSizeKb": 512
}

---

🚀 Quick Start

1. Install the extension from VS Code Marketplace
2. Run AI Memory: Start Service to begin indexing
3. Ask questions using AI Memory: Ask a Question
4. Visualize your project with AI Memory: Show Dependency Graph

---

🎯 Use Cases

• AI Agent Integration: Enable Claude Code and other agents to understand your codebase instantly
• Large Codebase Navigation: Quickly understand dependencies and architecture
• Code Impact Analysis: See which files are affected by changes before committing
• Token-Efficient AI Interactions: Get better answers within token limits
• Knowledge Base Creation: Build persistent project memory for team onboarding

---

🔧 Development

# Compile TypeScript
npm run compile

# Run tests
npm run test

# Watch for changes
npm run watch

# Start MCP server
npm run mcp

# Build for production
npm run package

---

📋 Requirements

• VS Code 1.80.0 or later
• Node.js 16+
• For AI features: Gemini API key (free tier available)

---

📄 License

MIT License - See LICENSE file for details

---

👨‍💻 Author

Developed with ❤️ by Pggamerbgoy

Designed to make coding faster, smarter, and token-efficient.

---

🤝 Contributing

Contributions welcome! Feel free to open issues and pull requests.

---

📚 Related Topics

vscode-extension · ai-agent · code-intelligence · semantic-search · ast-parsing · vector-database · context-caching · mcp · claude-code · workspace-analysis