Skip to main content
Claude Code is Anthropic’s flagship AI coding assistant, fully integrated with Maestro for powerful collaborative development.

​
Overview

Claude Code (claude) provides intelligent code generation, debugging, and refactoring capabilities with deep codebase understanding.
Claude Code is the most feature-complete agent in Maestro, supporting session resume, plan mode, image input, and comprehensive usage statistics.

​
Installation

1

Install Claude Code

Install via npm:
npm install -g @anthropic-ai/claude-code
2

Authenticate

Run Claude Code once to authenticate:
claude
Follow the prompts to sign in with your Anthropic account.
3

Verify in Maestro

Launch Maestro and create a new session with Claude Code as the agent type.

​
Capabilities

Claude Code provides the most comprehensive feature set of any Maestro-supported agent:
​
supportsResume
boolean
default:"true"
Resume existing sessions with --resume flag
​
supportsReadOnlyMode
boolean
default:"true"
Plan mode with --permission-mode plan (read-only access)
​
supportsJsonOutput
boolean
default:"true"
Structured JSON output with --output-format stream-json
​
supportsSessionId
boolean
default:"true"
Persistent session IDs for conversation continuity
​
supportsImageInput
boolean
default:"true"
Attach screenshots, diagrams, and images
​
supportsImageInputOnResume
boolean
default:"true"
Send images via --input-format stream-json on resumed sessions
​
supportsSlashCommands
boolean
default:"true"
Built-in commands like /help, /compact, /reset
​
supportsSessionStorage
boolean
default:"true"
Session history stored in ~/.claude/projects/
​
supportsCostTracking
boolean
default:"true"
Real-time cost information in USD
​
supportsUsageStats
boolean
default:"true"
Token counts for input and output
​
supportsThinkingDisplay
boolean
default:"true"
Streaming reasoning/thinking content
​
supportsContextMerge
boolean
default:"true"
Receive merged context from other sessions
​
supportsContextExport
boolean
default:"true"
Export conversation context for transfer
Source: src/main/agents/capabilities.ts:121

​
Command-Line Arguments

Maestro uses these base arguments for all Claude Code sessions: 
claude --print --verbose --output-format stream-json --dangerously-skip-permissions
​
--print
flag
Run in batch/headless mode (no interactive TUI)
​
--verbose
flag
Enable detailed logging output
​
--output-format
string
default:"stream-json"
Output format for structured parsing
​
--dangerously-skip-permissions
flag
Auto-approve all file operations (required by Maestro)

​
Resume Mode

To resume an existing session: 
claude --resume {sessionId} [additional args]
​
--resume
string
Claude session ID to resume

​
Plan Mode (Read-Only)

For read-only analysis without file modifications: 
claude --permission-mode plan [additional args]
​
--permission-mode
string
default:"plan"
Set to plan for read-only mode
Source: src/main/agents/definitions.ts:125

​
Session Storage

Claude Code stores session history in: 
~/.claude/projects/{project-name}/{session-id}/
Each session directory contains:
  • conversation.json - Full conversation history
  • metadata.json - Session metadata and configuration
  • files/ - Created/modified files during the session
Maestro can import existing Claude sessions for seamless continuation. Source: src/main/storage/claude-code-session-storage.ts

​
Output Format

Claude Code outputs newline-delimited JSON events: 
{
  "type": "system",
  "subtype": "init",
  "session_id": "abc123",
  "slash_commands": ["/help", "/compact", "/reset"]
}
Source: src/main/parsers/claude-code-parser.ts

​
Slash Commands

Claude Code supports these built-in commands:
Display available slash commands and usage information
Switch to compact output mode (less verbose)
Clear conversation history and start fresh
Show current context window usage

​
Error Patterns

Common error patterns Maestro detects:
​
API_KEY_INVALID
error
Pattern: Invalid API keySolution: Re-authenticate with claude command
​
RATE_LIMIT_EXCEEDED
error
Pattern: rate limit exceededSolution: Wait before retrying or upgrade your Anthropic plan
​
CONTEXT_LENGTH_EXCEEDED
error
Pattern: context length exceededSolution: Start a new session or use /reset to clear context
​
PERMISSION_DENIED
error
Pattern: permission deniedSolution: Check file permissions in working directory
Source: src/main/parsers/error-patterns.ts

​
Usage with Maestro Features

Auto Run

Fully supported - Claude Code works with playbooks and automated workflows

Group Chat

Participate in multi-agent group chats with other providers

Context Grooming

Export and merge conversation context between sessions

Session Discovery

Import existing Claude sessions from ~/.claude/projects/

​
Context Window

Claude Code uses Claude 3.5 Sonnet or Claude 4 models with:
  • Context Window: 200,000 tokens
  • Output Limit: ~8,000 tokens per response
Source: src/renderer/utils/contextUsage.ts:15

​
Best Practices

1

Use Descriptive Session Names

Name sessions based on the task for easier identification in history
2

Resume Instead of Restart

Use session resume to maintain context and reduce token usage
3

Enable Plan Mode for Analysis

Use read-only mode when you only need code analysis without modifications
4

Monitor Token Usage

Check the usage dashboard to track costs and optimize prompts

​
Troubleshooting

Verify installation:
which claude
claude --version
If not found, reinstall:
npm install -g @anthropic-ai/claude-code
Re-authenticate:
claude logout
claude
Check session exists:
ls ~/.claude/projects/
Session IDs are case-sensitive and must match exactly.

Build docs developers (and LLMs) love

Get started for freeTalk to us