🚀 Kade 3.0 - The Next Iteration of Agentic Coding
Kade has gone open source! Feel free to contribute, open issues, and star the GitHub repo.
Kade 3.0 is here
Completely redesigned, heavily upgraded, and built as the next major evolution of Kade. This release overhauls basically every aspect of the UI, improves the overall workflow from top to bottom, introduces a new Agent Manager page, and delivers more than a thousand fixes across the product.
This is not your average coding assistant. Kade is a revolutionary fork of Cline/RooCode/KiloCode designed to become the most powerful agentic IDE on the market - with unlimited rate limits, beautiful visual interfaces, and zero CLI bullshit.
Contact: For business inquiries, support, or other matters, please email support@kadei.org or visit kadei.org
💡 Why Kade Wins
Kade isn't just another coding assistant—it's a complete paradigm shift in how software is built. While others focus on simple chat interfaces, Kade provides a high-performance, agentic environment designed for professional engineers.
✅ Universal compatibility - Works with any AI, any provider
✅ Advanced context management - AI never gets confused
✅ Precision editing - Surgical code modifications
✅ Unlimited web access - Real-time information retrieval
✅ Enterprise authentication - Seamless integration
✅ Revolutionary sub-agent system - Delegate specialized tasks to dedicated AI agents
✅ Intelligent chat memory - Every conversation remembers its AI model across sessions
✅ Beautiful design - Joy to use every day
Kade AI in Action — High-Speed Agentic Workflows
The Kade Homescreen — Engineered for Speed and Precision
🤖 Revolutionary Sub-Agent System
The world's first true multi-agent IDE Chat
Forget legacy CLI tools. Kade brings true orchestration to your editor.
Kade introduces a groundbreaking sub-agent architecture that fundamentally changes how you work with AI. This is NOT your typical CLI-based sub-agent system - this is the first implementation that brings sub-agents directly into a rich, interactive IDE interface. Instead of one monolithic assistant trying to do everything, you can now delegate specialized tasks to dedicated agents with different capabilities, models, and contexts - each with their own dedicated chat interface.
How It Works
- Infinite Sub-Agent Spawning - Your main agent can spawn virtually unlimited sub-agents at once (yes, you could theoretically run 10,000+ sub-agents if you have a supercomputer and a very understanding cloud provider bill)
- Recursive Agent Creation - Sub-agents can spawn their own sub-agents, creating complex hierarchical task trees
- Dedicated Chat Interfaces - Each sub-agent gets its own chat tab in the IDE - you can literally go into your sub-agent's chat and talk directly to them
- Custom Model Selection - Choose ANY model from 50+ providers for each sub-agent - mix and match combinations like GPT-5 for architecture, Claude Opus for security, local Llama for documentation
- Isolated Contexts - Each sub-agent maintains its own conversation history, file context, and memory
- Seamless Coordination - Results flow back to the main agent automatically, or work directly with sub-agents in their dedicated chats
- Persistent Sub-Agent Sessions - Sub-agents remember their conversations even when you switch between them
Unprecedented Flexibility
- 50+ Provider Support - Mix OpenAI, Anthropic, Google, local models, and 45+ other providers in ANY combination
- Task-Specific Specialization - Create a "Testing Agent" with Gemini, "Security Agent" with Claude Opus, "Documentation Agent" with local Llama
- Direct Sub-Agent Interaction - Jump into any sub-agent's chat to give them specific instructions or review their work
- Cross-Agent Learning - Sub-agents can share insights and coordinate with each other through the main agent
- Hierarchical Task Trees - Build complex multi-level agent networks where sub-agents delegate to their own sub-agents
- Infinite Scalability - Scale from 2 agents to 200+ agents depending on your project complexity and budget
Real-World Use Cases
Main Agent: "I need to refactor this authentication system"
↳ Sub-Agent 1 (Security Specialist - Claude Opus): Analyzes vulnerabilities
↳ Sub-Agent 2 (Testing Expert - Gemini): Writes comprehensive tests
↳ You: Jump directly into Security Agent's chat to discuss specific threats
↳ Main Agent: Integrates all results and implements the refactor
This is how professional development teams actually work - and now your AI can too, with the freedom to choose the perfect model for each specialized task and interact with each agent directly in their own dedicated environment.
No other IDE Chat offers this level of multi-agent orchestration and model flexibility.
🧠 AI Chat Memory
Smart Model Persistence
Stop wasting time reconfiguring your workspace. Kade remembers your preferences so you can stay in the flow.
Kade's intelligent chat memory system ensures that every conversation remembers which AI model it was using and automatically restores that selection when you return. No more manually switching between GPT-4, Claude, or your local models every time you open a chat.
Key Benefits
- Persistent Model Selection - Each chat window maintains its own AI model choice across sessions
- Multi-Model Workflows - Run GPT-5 in one chat, Claude in another, Gemini in a third - simultaneously
- Seamless Context Switching - Jump between conversations without losing your model preferences
- Zero Configuration - Set it once, it remembers forever
Perfect For
- Specialized Workflows - Keep a Claude chat for writing, GPT-5 for coding, local model for quick tasks
- Model Comparison - Test the same prompt across different models in parallel windows
- Team Collaboration - Share chat sessions with model preferences intact
- Long-Running Projects - Return to month-old conversations with the exact same AI setup
Work smarter, not harder - let Kade remember the details while you focus on building.
The death of JSON/XML - one syntax to rule them all:
While the AI industry defaulted to JSON and XML for tool calling, these legacy formats are deeply flawed. Native JSON function calling suffers from severe cross-provider compatibility issues, strict schema validation failures, and massive token overhead. XML is notoriously janky for LLMs to generate cleanly, often leading to broken tags, whitespace issues, and escaped character nightmares.
Kade solves this by introducing two revolutionary text-based protocols that leverage what AI models are already perfectly trained on: Markdown (code blocks) and Unified (CLI syntax). By speaking the model's native language, we eliminate syntax errors, drop provider lock-in entirely, and drastically reduce token usage.
When to use each:
- Unified is tailored for maximum accuracy and precision, offering a concise CLI-like syntax that eliminates ambiguity and ensures reliable tool execution.
- Markdown is optimized for speed and rapid development, leveraging familiar code block structure that models can generate quickly and naturally.
- Native JSON & XML remain fully supported for legacy integrations and specific provider requirements.
Kade is the only agent that gives you the freedom to choose, supporting all four protocols out of the box.
# JSON Hell (15+ lines of nested chaos):
{
"tool_calls": [
{
"id": "call_abc123",
"type": "function",
"function": {
"name": "read",
"arguments": "{\"path\": \"src/main.ts\", \"lines\": \"1-50\"}"
}
}
]
}
# Unified Simplicity (1 line):
<<read --path src/main.ts --lines 1-50>>
# Markdown Elegance (Native to AI):
```read
src/main.ts --lines 1-50
Revolutionary Features
- Works with EVERY provider - OpenAI, Anthropic, Google, Local Models
- Zero provider lock-in - Pure text protocol, no native function calling
- Streaming-safe - Real-time AI responses handled gracefully
- Multi-tool batching - Multiple operations in one turn
- (N) Limit - Set a limit on how many tool calls can be used in one turn, set it to unlimited or disable batch tool-calling altogether.
- Bulletproof reliability - Escape-safe, error-resilient, debug-friendly
🧠 Smart Context Engineering
Revolutionary context management that never gets stale:
- Turn-by-Turn Refresh - Files automatically refresh after every edit
- Smart Edit Tracking - "Edit #N" progression shows change history
- Token Optimization - Context management system only keeps latest versions
- Persistent Memory - Restores context on restart with cross-platform support
⏮️ Perfect Undo/Redo System
VS Code snapshots with zero context poisoning:
- Instant Revert - Undo any AI edit with one click, no token waste
- Smart Snapshots - Captures file states before every tool execution
- Context-Safe - Works perfectly with context management system (no stale context)
- A/B Testing - Toggle between undos/redos to compare approaches
- Selective Recovery - Keep 9 good edits, revert just the 1 bad one
How It Works
- Pre-edit snapshots of all files before AI tools run
- One-click undo restores original content instantly
- Redo capability re-applies changes when needed
- File creation/deletion handled automatically
- Zero context corruption - Context management system keeps everything in sync
Perfect for: Fixing bad edits, testing variations, recovering from mistakes without losing thousands of tokens.
🔧 Insane Edit Precision
Engineering that makes other editors look like toys:
- Whitespace Chaos Normalization - Handles completely wrong indentation/spacing
- Token-based Matching - Intelligent fuzzy matching with similarity scoring
- Multi-block Edits - Surgical precision with multiple simultaneous changes
- Line Range Support - Targeted modifications
[start, end]
- Unicode & Cross-platform - Works with any character set and line endings
🌐 Unlimited Web Intelligence
Custom web system with no API limits:
- Startpage-powered search - Real-time results from the web
- Content extraction - Any URL converted to clean markdown
- Unlimited queries - No per-day limits, completely local implementation
- Intelligent filtering - HTML parsing, result ranking, error handling
LobeHub API integration gives you instant access:
- 25,000+ MCPs - One-click installation from extension
- Universal Schema Support - Unified, Markdown, JSON, XML, Native Provider all work
- Real-time Marketplace - Search, install, activate without leaving VS Code
- Enterprise Features - Security scanning, team sharing, custom repositories
🔐 Enterprise Authentication
Complete OAuth support for all major platforms:
- Antigravity (Google Gemini Code Assist)
- Kilo Code • Gemini CLI • Claude Code • OpenAI Codex
- Enterprise-grade - Secure token storage, auto-refresh, multi-step onboarding
- One-click auth - Local callback servers with CSRF protection
🎨 Beautiful by Design
Next-generation UI built for joy:
- Modern components - React library with responsive design
- Enhanced UX - Streamlined chat, better file management, real-time collaboration
- Performance optimized - Lazy loading, memory optimization, background processing
🏗️ Champion Architecture
Built to surpass every limitation:
- Monorepo excellence - Turbo-powered builds, TypeScript everywhere
- Performance first - Handles massive codebases with intelligent caching
- Developer experience - Hot reload, comprehensive testing, debug tools
🗺️ Roadmap - What's Coming Next
The revolution is just getting started:
VS Code Fork [90% Complete]
ETA: Q2 2026
- Native AI Core - Rebuilt from the ground up, moving beyond extension limitations for deep IDE integration.
- Zero-Latency UI - A completely redesigned interface optimized for agentic workflows.
- Full System Control - Direct access to the IDE's internal APIs for unprecedented automation.
Kade CUA (Computer Use Agent) [95% Complete]
ETA: < 1 Month
- Most Advanced Computer Use Agent for every provider
- Universal CUA that works with OpenAI, Anthropic, Google, Local Models
- No provider lock-in - one agent to rule them all
- Screen understanding, mouse/keyboard control, app automation
Future Vision
- Multi-language support - Beyond TypeScript/JavaScript
- Team collaboration - Shared AI sessions and knowledge bases
- Enterprise features - Advanced security, compliance, SSO
- Mobile development - iOS/Android AI assistance
- Cloud deployment - AI-powered CI/CD and infrastructure management
Kade AI Agent represents a fundamental shift in how developers interact with AI. Born from a mission to transform existing coding assistants into truly revolutionary tools, Kade has evolved beyond simply catching up to competitors—it's setting entirely new standards for what's possible.
🚀 Built for the Future
Architecture: Extensible, performance-optimized, cross-platform solution engineered for enterprise scale.
Kade AI Agent is redefining the boundaries of development tools. Every feature is meticulously engineered to solve real-world challenges developers face daily, delivering unprecedented productivity and capability.
This isn't just an upgrade—it's a revolution. Welcome to the future of coding.
Contact: For business inquiries, support, or other matters, please email support@kadei.org or visit kadei.org
Built with ❤️ and a healthy dose of engineering insanity.
