AgentBuilder Documentation
Visual IDE for building Claude AI agents with drag-and-drop simplicity. Think Blender for agent workflows — visual node graph, real-time testing, export to production config.Getting Started
New to AgentBuilder? Start here to get up and running quickly.Installation & Setup
Get AgentBuilder running on your machine in 5 minutes
Quick Start
Build a working “Email Summarizer” agent step-by-step in 10 minutes
Workspace Basics
Understand projects, folders, and organization
Core Concepts
Master the fundamentals of building AI agents visually.Node Types Reference
Learn about the 7 node types that make up your agent systems:Agent Node
Main agent configuration: system prompt, model, permissions
Skills Node
Agent capabilities: file operations, API calls, code execution
Memory Node
Persistent knowledge: conversation history, knowledge bases
Commands Node
Custom slash commands:
/summarize, /review-prMCPs Node
External tool integrations: Context7 docs, databases, APIs
Hooks Node
Lifecycle event handlers: PreToolUse, PostToolUse, SessionStart
Sub-agents Node
Delegated specialized agents: code reviewer, researcher
Node Types Overview
Complete reference for all node types
What You Can Build
AgentBuilder enables you to create sophisticated AI agent systems:Single-Purpose Agents
Single-Purpose Agents
- Email handler and summarizer
- Code reviewer with custom rules
- Document analyzer and reporter
- Customer support bot
Multi-Agent Systems
Multi-Agent Systems
- Research coordinator with specialist sub-agents
- Code generation pipeline (planner → writer → reviewer)
- Customer service orchestrator (routing → specialist → escalation)
Complex Workflows
Complex Workflows
- Agents with custom lifecycle hooks
- Memory-enabled conversational agents
- Tool-augmented agents with MCP integrations
- Command-driven automation bots
Architecture Overview
- Electron 28 - Desktop application framework
- Vue 3 - Modern UI framework with TypeScript
- Vue Flow - Node graph canvas
- Pinia - State management
- Claude Agent SDK - AI agent execution
Why AgentBuilder?
The Problem
Manually editing.claude configuration folders is error-prone and difficult to visualize:
- Hard to understand agent architecture at a glance
- Easy to misconfigure permissions and tool access
- Difficult to manage complex multi-agent systems
- No way to test changes in real-time
The Solution
AgentBuilder provides a visual interface for building Claude AI agents: Visual Node Graph- Drag-and-drop canvas for building agent workflows
- Clear visualization of agent architecture
- Real-time validation and error checking
- Chat interface built right into the IDE
- Test agents as you build them
- See agent thinking process and tool calls
- One-click export to
.claudefolder structure - Generates all configuration files automatically
- Compatible with Claude Code CLI
- Skills, Commands, MCPs, Sub-agents, Hooks
- All managed through intuitive GUI
- Reusable component library
Quick Links
Your First Agent
Build your first agent in 10 minutes
Installation & Setup
Get AgentBuilder installed in 5 minutes
Node Types Reference
Learn about all 7 node types
Claude Code Setup
Configure Claude Code for your workflow
System Requirements
- Node.js: ≥18.19.9
- Platform: macOS, Windows, or Linux
- RAM: 4GB minimum, 8GB recommended
- Disk Space: 500MB for app + workspace storage
Community & Support
Report a Bug
Found an issue? Let us know on GitHub
Join Discord
Get help and connect with the community
Contribute
Help improve AgentBuilder