Skip to main content

​
Workspace Basics

Learn how ClaudeSpace organizes your agents, projects, and workspaces.

​
Workspace Hierarchy

ClaudeSpace uses a three-level organization system: 
📁 My-Agents-Workspace/          ← Workspace (root folder)
├── 📁 email-summarizer/          ← Project folder
│   ├── workflow.json             ← Visual graph data
│   ├── agent.json                ← Agent metadata
│   ├── .claude/                  ← Claude Code config
│   │   ├── CLAUDE.md             ← System prompt
│   │   ├── settings.json         ← Agent definitions
│   │   ├── settings.local.json   ← Runtime settings
│   │   ├── skills/               ← Skill implementations
│   │   └── memory/               ← Memory files
│   └── context/                  ← Context files
├── 📁 code-reviewer/             ← Another project
│   └── ...
└── 📁 research-assistant/        ← Another project
    └── ...

​
Project

A project is a single agent system with its visual workflow and configuration. Contains:
  • Visual workflow - Node graph (canvas state)
  • Agent configuration - System prompt, settings, permissions
  • Skills - Custom capabilities and tools
  • Memory - Knowledge base files
  • Commands - Slash command definitions
  • Hooks - Lifecycle event handlers
  • Sub-agents - Specialized delegated agents
  • Context files - Documents and references
Lifecycle:
  1. Created - New project initialized with main agent card
  2. Developed - Nodes added, configured, connected
  3. Tested - Agent run via chat interface
  4. Exported - .claude folder generated for production
  5. Deployed - Used with Claude Code CLI
Metadata stored in agent.json: 
{
  "name": "Email Summarizer",
  "created": "2025-01-10T10:30:00Z",
  "modified": "2025-01-10T15:45:00Z",
  "iconImage": "icons/email.png",
  "description": "Analyzes emails and suggests responses"
}

​
Node Library

The Node Library contains reusable components shared across all projects in a workspace. Location: <workspace>/.library/ Contains:
  • Skills - Custom tools and capabilities
  • MCP Servers - External tool integrations
  • Hooks - Lifecycle event scripts
  • Commands - Reusable slash commands
  • Sub-agents - Shared agent templates
  • Memory Templates - Knowledge base structures
How it works:
  1. Create components in Settings → Node Managers
  2. Components saved to workspace library
  3. Drag from sidebar to add to any project
  4. Each project gets its own instance (can customize)
  5. Edit library item → all projects can use updated version

​
Customizing Workspaces

Access: Settings (⚙️) → General Tab → Workspace Customization Customize:
  • Icon - Emoji or image for visual identification
  • Title - Display name (different from folder name)
  • Description - Purpose or category notes

​
Delete Project

Warning: Deletion is permanent and cannot be undone. Steps:
  1. Right-click project card on dashboard
  2. Select “Delete Project” from context menu
  3. Confirmation dialog appears
  4. Type project name to confirm
  5. Click “Delete Permanently”
[Screenshot: Delete confirmation dialog] What gets deleted:
  • All project files (workflow.json, agent.json)
  • .claude/ configuration folder
  • Skills, memory, commands, hooks
  • Cannot be recovered
Best practice: Export project before deleting to keep backup.
agent.json - Project Metadata 
{
  "name": "Email Summarizer",
  "description": "Analyzes emails and suggests responses",
  "created": "2025-01-10T10:30:00Z",
  "modified": "2025-01-10T15:45:00Z",
  "iconImage": "data:image/png;base64,..."
}
Purpose: Dashboard display, search, sorting
.claude/ Folder Structure Generated when you export or test agent: 
.claude/
├── CLAUDE.md                 # System prompt (main agent)
├── settings.json             # Agent definitions
├── settings.local.json       # Runtime config
├── .mcp.json                 # MCP servers (at project root)
├── agents/                   # Sub-agent definitions
│   ├── code-reviewer.md
│   └── researcher.md
├── skills/                   # Skill implementations
│   ├── file-system/
│   │   ├── SKILL.md
│   │   └── scripts/
│   └── github-api/
│       ├── SKILL.md
│       └── references/
├── memory/                   # Knowledge base files
│   ├── company-info.md
│   └── team-contacts.md
├── commands/                 # Slash command definitions
│   ├── summarize.md
│   └── review.md
└── hooks/                    # Lifecycle event scripts
    ├── pre-tool-use.sh
    └── post-tool-use.sh
Purpose: Claude Code CLI compatibility, version control

​
Next Steps

Continue learning: Build more agents:
Ready to build complex agent systems?Explore Node Types →