guIDE：为AI时代重新设计的原生AI代码编辑器，本地LLM与52+ MCP工具深度集成

guIDE is a code editor built from scratch for the AI era, integrating local LLM inference engine, 52+ MCP tools, browser automation, and RAG code intelligence. It supports 9 cloud AI providers, aiming to deliver a true AI-native development experience while addressing limitations of traditional editors like VS Code with AI plugins (e.g., privacy concerns, no local model support, limited context).

VS Code, though popular, was born before the AI era. AI assistants like GitHub Copilot are plugins, leading to issues:

• Code sent to cloud APIs (privacy risks)
• No native local model support
• Limited tool execution ability (AI can only suggest, not act)
• Context restricted to open files (no full codebase awareness) guIDE was created to break these constraints.

guIDE's design follows three principles: Local-first, AI-native, Full Control. Comparison with VS Code + Plugins:

Feature	VS Code + Plugins	guIDE
AI Integration	Plugin extension	Built into core architecture
Local LLM Support	No native	node-llama-cpp + CUDA/Vulkan acceleration
Data Privacy	Code sent to cloud	Fully local, no telemetry
Tool Calls	Independent cloud API calls	52 MCP tools for AI autonomous use
Browser Automation	Requires extra tools	Built-in BrowserView (15 tools)
Cost	Monthly $10-20	Free local models, pay-as-you-go for cloud
Context Scope	Open files only	RAG-indexed entire codebase + persistent memory

Local LLM Inference Engine

• Supports any GGUF models (Qwen, Llama, Mistral, DeepSeek)
• GPU acceleration (CUDA/Vulkan) with layer offloading
• Adaptive context window (max 32K tokens based on memory)
• Flash Attention for efficient memory use

MCP Tool Ecosystem

MCP (Model Context Protocol) enables AI-environment interaction with 52 tools:

• File operations (read/write/search/analyze)
• Terminal control (commands, process management)
• Browser automation (15 tools)
• Git integration
• Persistent memory

RAG-Driven Code Intelligence

• Auto builds code vector index
• Semantic search for intent understanding
• Injects relevant code snippets into context
• Analyzes cross-file references

Familiar Editing Experience

Uses Monaco Editor (same as VS Code):

• Syntax highlighting, IntelliSense, multi-cursor
• Command panel, split editing, themes/icons

AI-Enhanced Features

• Inline chat (Ctrl+I for code modification)
• Next edit suggestion (Tab to accept)
• Multi-agent mode (/agent for background tasks)
• Plan mode (AI makes plan before execution)
• Foldable reasoning for thought-chain models

Context Management

Priority-based prompt assembly: Memory → Tools → Errors → RAG → Code → Web Search

• Persistent cross-session memory
• Custom .prompt.md for project-specific context
• Token usage indicator (progress bar)

Cloud Provider Integration

Supports 9 cloud AI providers (OpenAI, Anthropic, Google, etc.) with:

• Pre-configured free tiers
• Flexible switch between local/cloud models
• Pay-as-you-go, no mandatory subscriptions

Privacy & Security

• Zero telemetry (no usage data upload)
• Zero cloud dependency for core functions
• Code never leaves the machine (local inference)
• Suitable for confidential environments (finance, military)

Open Source & Customization

• Source-available (modify prompts, tools, UI)
• Custom system prompts to fine-tune AI behavior
• Add new MCP tools
• Full control over tech stack

Target Users

• Privacy-sensitive scenarios (proprietary code, compliance industries)
• Offline environments (network-limited areas)
• Cost-conscious developers (avoid subscriptions)
• Model enthusiasts (try open-source models)
• High-security organizations (government, military)

Conclusion

guIDE redefines AI code editors by building from scratch for the AI era, not as a plugin. Its core values:

• Your code, your model, your machine
• No subscriptions, no internet needed, no privacy worries
• AI understands the entire project, not just fragments

As local LLMs and hardware improve, guIDE-like local-first tools will likely gain more traction.