Lifecycle Events

Claude Code provides lifecycle hooks for various events during session execution, tool usage, and system operations.

Session Lifecycle

SessionStart

Runs when a new session starts or resumes. Timing: After session initialization, before first user message Configuration:

{
  "hooks": {
    "SessionStart": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "bash ${CLAUDE_PLUGIN_ROOT}/setup-session.sh"
          }
        ]
      }
    ]
  }
}

Input:

{
  "session_id": "abc123",
  "working_directory": "/path/to/project",
  "is_resume": false
}

Use cases:

Environment configuration
Loading project-specific context
Initializing external services
Setting up Docker containers
Loading API credentials

SessionStart hooks are deferred during startup to reduce time-to-interactive by ~500ms.

Example:

setup-session.sh

#!/bin/bash
# Configure environment for Claude session

export PROJECT_ENV="development"
source .env.local 2>/dev/null || true

# Start development services
docker-compose up -d db redis

echo "Session environment configured" >&2
exit 0

Stop

Runs when a session is ending (user exit or interruption). Timing: Before session cleanup and exit Configuration:

{
  "hooks": {
    "Stop": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "${CLAUDE_PLUGIN_ROOT}/hooks/stop-hook.sh"
          }
        ]
      }
    ]
  }
}

Input:

{
  "session_id": "abc123",
  "transcript_path": "/path/to/transcript.jsonl",
  "last_assistant_message": "Final message from Claude",
  "working_directory": "/path/to/project"
}

Output (optional):

{
  "decision": "block",
  "reason": "Continue with new prompt",
  "systemMessage": "Loop iteration 5"
}

Stop hooks can block session exit by returning "decision": "block". This enables continuous loop patterns.

Use cases:

Cleanup operations
Saving session state
Stopping services
Creating session summaries
Continuous loop workflows (Ralph Wiggum pattern)

Example (Ralph Wiggum Loop):

stop-hook.sh

#!/bin/bash
# Prevent exit and continue loop

set -euo pipefail

# Read hook input
HOOK_INPUT=$(cat)

# Check if loop is active
if [[ ! -f ".claude/ralph-loop.local.md" ]]; then
  exit 0  # Allow exit
fi

# Parse state
FRONTMATTER=$(sed -n '/^---$/,/^---$/{ /^---$/d; p; }' ".claude/ralph-loop.local.md")
ITERATION=$(echo "$FRONTMATTER" | grep '^iteration:' | sed 's/iteration: *//')
MAX_ITERATIONS=$(echo "$FRONTMATTER" | grep '^max_iterations:' | sed 's/max_iterations: *//')

# Check if max reached
if [[ $MAX_ITERATIONS -gt 0 ]] && [[ $ITERATION -ge $MAX_ITERATIONS ]]; then
  rm ".claude/ralph-loop.local.md"
  exit 0  # Allow exit
fi

# Continue loop
NEXT_ITERATION=$((ITERATION + 1))
PROMPT_TEXT=$(awk '/^---$/{i++; next} i>=2' ".claude/ralph-loop.local.md")

# Update state
sed "s/^iteration: .*/iteration: $NEXT_ITERATION/" ".claude/ralph-loop.local.md" > tmp
mv tmp ".claude/ralph-loop.local.md"

# Block exit and feed prompt back
jq -n \
  --arg prompt "$PROMPT_TEXT" \
  --arg msg "🔄 Ralph iteration $NEXT_ITERATION" \
  '{
    "decision": "block",
    "reason": $prompt,
    "systemMessage": $msg
  }'

exit 0

Setup

Runs for repository setup and maintenance operations. Timing: When invoked via CLI flags: --init, --init-only, or --maintenance Configuration:

{
  "hooks": {
    "Setup": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "bash ./scripts/setup.sh"
          }
        ]
      }
    ]
  }
}

Input:

{
  "mode": "init",
  "working_directory": "/path/to/project"
}

Use cases:

Repository initialization
Dependency installation
Database migrations
Environment setup
Periodic maintenance tasks

Example:

# Run setup on new clone
claude --init

# Run setup without starting session
claude --init-only

# Run maintenance tasks
claude --maintenance

Tool Lifecycle

PreToolUse

Runs before a tool is executed. Timing: After Claude requests tool use, before execution Configuration:

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash",
        "hooks": [
          {
            "type": "command",
            "command": "python3 /path/to/validator.py"
          }
        ]
      }
    ]
  }
}

Input:

{
  "session_id": "abc123",
  "tool_name": "Bash",
  "tool_input": {
    "command": "git push origin main",
    "workdir": "/path/to/project"
  }
}

Output:

{
  "decision": "block",
  "additionalContext": "This command requires manual approval"
}

Exit codes:

0 - Allow execution
1 - Show stderr to user only (allow execution)
2 - Block execution, show stderr to Claude

Use cases:

Command validation
Security checks
Command transformation
Rate limiting
Audit logging

Example (Command Validator):

validator.py

#!/usr/bin/env python3
import json
import sys
import re

VALIDATION_RULES = [
    (
        r"^grep\b(?!.*\|)",
        "Use 'rg' (ripgrep) instead of 'grep' for better performance",
    ),
    (
        r"^find\s+\S+\s+-name\b",
        "Use 'rg --files -g pattern' instead of 'find -name'",
    ),
]

def validate_command(command: str) -> list[str]:
    issues = []
    for pattern, message in VALIDATION_RULES:
        if re.search(pattern, command):
            issues.append(message)
    return issues

try:
    data = json.load(sys.stdin)
except json.JSONDecodeError as e:
    print(f"Error: Invalid JSON: {e}", file=sys.stderr)
    sys.exit(1)

tool_name = data.get("tool_name", "")
if tool_name != "Bash":
    sys.exit(0)

command = data.get("tool_input", {}).get("command", "")
if not command:
    sys.exit(0)

issues = validate_command(command)
if issues:
    for msg in issues:
        print(f"• {msg}", file=sys.stderr)
    sys.exit(2)  # Block execution

sys.exit(0)  # Allow execution

PostToolUse

Runs after a tool completes execution. Timing: After tool execution, before result is sent to Claude Configuration:

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Bash",
        "hooks": [
          {
            "type": "command",
            "command": "python3 /path/to/logger.py"
          }
        ]
      }
    ]
  }
}

Input:

{
  "session_id": "abc123",
  "tool_name": "Bash",
  "tool_input": {
    "command": "npm test"
  },
  "tool_output": {
    "stdout": "All tests passed",
    "stderr": "",
    "exit_code": 0
  }
}

Use cases:

Result logging
Metrics collection
Notification triggers
Result transformation
Error analysis

Git Worktree Lifecycle

WorktreeCreate

Runs when a git worktree is created for agent isolation. Timing: After worktree creation, before agent starts Configuration:

{
  "hooks": {
    "WorktreeCreate": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "bash ./scripts/setup-worktree.sh"
          }
        ]
      }
    ]
  }
}

Input:

{
  "worktree_path": "/path/to/worktree",
  "branch_name": "claude/task-123",
  "base_branch": "main"
}

Use cases:

Worktree-specific configuration
Installing dependencies
Database setup
Environment configuration

WorktreeRemove

Runs when a git worktree is being removed. Timing: Before worktree deletion Configuration:

{
  "hooks": {
    "WorktreeRemove": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "bash ./scripts/cleanup-worktree.sh"
          }
        ]
      }
    ]
  }
}

Input:

{
  "worktree_path": "/path/to/worktree",
  "branch_name": "claude/task-123"
}

Use cases:

Cleanup operations
Stopping services
Collecting artifacts
Backup operations

Configuration Lifecycle

ConfigChange

Runs when configuration files change during a session. Timing: When settings.json or related config files are modified Configuration:

{
  "hooks": {
    "ConfigChange": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "python3 /path/to/audit-config.py"
          }
        ]
      }
    ]
  }
}

Input:

{
  "session_id": "abc123",
  "config_file": "/path/to/settings.json",
  "changes": {
    "permissions.allow": ["Bash"]
  }
}

Output (optional):

{
  "decision": "block",
  "reason": "Configuration changes require administrator approval"
}

Use cases:

Enterprise security auditing
Configuration validation
Change approval workflows
Compliance enforcement

Agent Team Lifecycle

TeammateIdle

Runs when an agent teammate becomes idle. Timing: When teammate completes work and awaits instructions Configuration:

{
  "hooks": {
    "TeammateIdle": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "python3 /path/to/notify.py"
          }
        ]
      }
    ]
  }
}

Input:

{
  "session_id": "abc123",
  "agent_id": "teammate-1",
  "agent_type": "code-reviewer"
}

TaskCompleted

Runs when a background task completes. Timing: When task finishes successfully or with error Configuration:

{
  "hooks": {
    "TaskCompleted": [
      {
        "hooks": [
          {
            "type": "command",
            "command": "python3 /path/to/task-logger.py"
          }
        ]
      }
    ]
  }
}

Input:

{
  "session_id": "abc123",
  "task_id": "task-1",
  "status": "success",
  "result": "Task completed successfully",
  "duration_ms": 5432
}

SubagentStop

Runs when a subagent session ends. Timing: After subagent completes, before returning to parent Input:

{
  "session_id": "abc123",
  "parent_session_id": "parent-123",
  "agent_type": "code-analyzer",
  "last_assistant_message": "Analysis complete"
}

Best Practices

Keep hooks fast

Hooks run synchronously and block execution:

Target < 100ms execution time
Use background processes for slow work
Cache expensive operations
Avoid network calls when possible

Handle errors gracefully

try:
    data = json.load(sys.stdin)
    # Process data
except Exception as e:
    print(f"Hook error: {e}", file=sys.stderr)
    sys.exit(1)  # Show error to user, allow execution

Use appropriate exit codes

0 - Success, allow operation
1 - Non-blocking error (user sees stderr)
2 - Blocking error (Claude sees stderr)
Choose based on whether Claude should know about the issue

Validate input

if "tool_input" not in data:
    sys.exit(0)  # Gracefully handle missing data

if data.get("tool_name") != "Bash":
    sys.exit(0)  # Only process relevant tools

Clean up state

# Store session-specific state
STATE_FILE = f"~/.claude/hook-state-{session_id}.json"

# Clean up old state files
cleanup_old_states()  # Runs periodically

Next Steps

Hook Examples

See real-world hook implementations and patterns

Commands

Hooks

Configuration Files

Session Lifecycle

SessionStart

Stop

Setup

Tool Lifecycle

PreToolUse

PostToolUse

Git Worktree Lifecycle

WorktreeCreate

WorktreeRemove

Configuration Lifecycle

ConfigChange

Agent Team Lifecycle

TeammateIdle

TaskCompleted

SubagentStop

Best Practices

Next Steps

Hook Examples

Build docs developers (and LLMs) love

Commands

Hooks

Configuration Files

​Session Lifecycle

​SessionStart

​Stop

​Setup

​Tool Lifecycle

​PreToolUse

​PostToolUse

​Git Worktree Lifecycle

​WorktreeCreate

​WorktreeRemove

​Configuration Lifecycle

​ConfigChange

​Agent Team Lifecycle

​TeammateIdle

​TaskCompleted

​SubagentStop

​Best Practices

​Next Steps

Hook Examples

Build docs developers (and LLMs) love

Session Lifecycle

SessionStart

Stop

Setup

Tool Lifecycle

PreToolUse

PostToolUse

Git Worktree Lifecycle

WorktreeCreate

WorktreeRemove

Configuration Lifecycle

ConfigChange

Agent Team Lifecycle

TeammateIdle

TaskCompleted

SubagentStop

Best Practices

Next Steps