gwq - Git Worktree Manager

gwq is a CLI tool for efficiently managing Git worktrees. Like how ghq manages repository clones, gwq provides intuitive operations for creating, switching, and deleting worktrees using a fuzzy finder interface.

Why gwq?

Git worktrees allow you to check out multiple branches from the same repository into separate directories. This is particularly powerful when:

• Working on multiple features simultaneously
• Running parallel AI coding agents on different tasks
• Reviewing code while developing new features
• Testing changes without disrupting your main workspace

AI Coding Agent Workflows

One of the most powerful applications of gwq is enabling parallel AI coding workflows. Instead of having a single AI agent work sequentially through tasks, you can leverage multiple worktrees to have multiple AI agents (like Claude Code) work on different parts of your project simultaneously:

# Create worktrees for parallel development
gwq add feature/authentication
gwq add feature/data-visualization
gwq add bugfix/login-issue

# Each AI agent can work in its own worktree
cd ~/worktrees/myapp-feature-authentication && claude
cd ~/worktrees/myapp-feature-data-visualization && claude
cd ~/worktrees/myapp-bugfix-login-issue && claude

Since each worktree has its own working directory with isolated files, AI agents can work at full speed without waiting for each other's changes or dealing with merge conflicts. You can monitor all agent activity in real-time with gwq status --watch. This approach is ideal for:

• Independent tasks: Each AI agent focuses on a separate feature or component
• Parallel migrations: Multiple agents can migrate different parts of your codebase simultaneously
• Code review workflows: One agent writes code while another reviews it in a separate worktree
• Testing isolation: Run tests in one worktree while developing in another
• Progress monitoring: Track which agents have made changes and when using the status dashboard

Installation

Using Go

go install github.com/d-kuro/gwq/cmd/gwq@latest

From Source

git clone https://github.com/d-kuro/gwq.git
cd gwq
go build -o gwq ./cmd/gwq

Quick Start

# Create a new worktree with new branch
gwq add -b feature/new-ui

# List all worktrees
gwq list

# Check status of all worktrees
gwq status

# Get worktree path
gwq get

# Remove a worktree
gwq remove feature/old-ui

Features

• Fuzzy Finder Interface: Built-in fuzzy finder for intuitive branch and worktree selection
• Smart Navigation: Quick switching between worktrees with pattern matching
• Global Worktree Management: Access all your worktrees across repositories from anywhere
• Tab Completion: Full shell completion support for branches, worktrees, and configuration
• Configuration Management: Customize worktree directories and naming conventions
• Preview Support: See branch details and recent commits before selection
• Clean Operations: Automatic cleanup of deleted worktree information
• Branch Management: Optional branch deletion when removing worktrees
• Home Directory Display: Option to display paths with ~ instead of full home directory path
• Worktree Status Dashboard: Monitor all worktrees' git status, changes, and activity at a glance
• Tmux Session Management: Run and manage long-running processes in persistent tmux sessions
• AI Development Automation: Structured task queue system for automated development with Claude Code
• Task Management: Priority-based task scheduling with dependency resolution and resource management

Global Worktree Management

gwq automatically discovers all worktrees in your configured base directory, allowing you to access them from anywhere on your system:

• Outside Git Repositories: When you run gwq list outside a git repository, it automatically discovers and shows all worktrees in the configured base directory
• Inside Git Repositories: By default, shows only worktrees for the current repository. Use the -g flag to see all worktrees from the base directory
• Automatic Discovery: All worktrees in the base directory are automatically discovered, including those created with native git commands
• No Registry Required: Uses filesystem scanning instead of maintaining a separate registry file

This feature is particularly useful when:

• Managing multiple projects simultaneously
• Quickly jumping between different repositories' worktrees
• Getting an overview of all active development branches across projects

Note: All worktrees located in the configured base directory (default: ~/worktrees) are automatically discovered, regardless of how they were created.

Commands

gwq add

Create a new worktree

# Create worktree with new branch
gwq add -b feature/new-ui

# Create from existing branch
gwq add main  # Creates worktree from existing 'main' branch

# Create at specific path with new branch
gwq add -b feature/new-ui ~/projects/myapp-feature

# Create from remote branch
gwq add -b feature/api-v2 origin/feature/api-v2

# Interactive branch selection with fuzzy finder
gwq add -i

gwq list

Display all worktrees

# Simple list
gwq list
# Output:
# BRANCH        PATH
# ● main        ~/ghq/github.com/user/project
# feature/api   ~/worktrees/github.com/user/project/feature-api
# bugfix/login  ~/worktrees/github.com/user/project/bugfix-login

# Detailed information
gwq list -v

# JSON format for scripting
gwq list --json

# Show all worktrees from base directory (from anywhere)
gwq list -g

gwq get

Get worktree path

# Get path and change directory
cd $(gwq get feature)

# Get path for specific worktree
gwq get main

# Use with other commands
ls -la $(gwq get feature)

# Use null-terminated output with xargs
gwq get -0 feature | xargs -0 -I {} echo "Path: {}"

# Get path from global worktrees
gwq get -g myapp:feature

gwq exec

Execute command in worktree directory

# Run tests in feature branch
gwq exec feature -- npm test

# Pull latest changes in main branch
gwq exec main -- git pull

# Run multiple commands
gwq exec feature -- sh -c "git pull && npm install && npm test"

# Stay in worktree directory after command
gwq exec --stay feature -- npm install

# Execute in global worktree
gwq exec -g myapp:feature -- make build

# Interactive selection with fuzzy finder (when multiple matches)
gwq exec feature -- npm test  # Shows fuzzy finder if multiple feature/* branches exist

# Select from all worktrees with fuzzy finder
gwq exec -- npm test  # Shows all worktrees in fuzzy finder

gwq remove

Delete worktree

# Select and delete using fuzzy finder
gwq remove

# Delete by pattern
gwq remove feature/old

# Force delete
gwq remove -f feature/broken

# Delete worktree and branch together
gwq remove -b feature/completed

# Force delete branch even if not merged
gwq remove -b --force-delete-branch feature/abandoned

# Preview what would be deleted
gwq remove --dry-run -b feature/old

# Remove from any worktree in base directory (from anywhere)
gwq remove -g myapp:feature/old

Branch Deletion Options:

• By default, gwq remove only deletes the worktree directory, preserving the branch
• Use -b/--delete-branch to also delete the branch after removing the worktree
• The branch deletion uses safe mode (git branch -d) by default, which prevents deletion of unmerged branches
• Use --force-delete-branch with -b to force delete even unmerged branches (git branch -D)

gwq status

Monitor the status of all worktrees

# Show status in table format
gwq status
# Output:
# BRANCH          STATUS       CHANGES                   ACTIVITY
# ● main          changed      8 modified, 8 untracked   just now
#   feature/api   up to date   -                         2 hours ago
#   bugfix/login  changed      3 added, 2 modified       30 mins ago

# Watch mode - auto-refresh every 5 seconds
gwq status --watch

# JSON output for scripting
gwq status --json

# Show additional information (ahead/behind, processes)
gwq status --verbose

# Filter by status
gwq status --filter changed
gwq status --filter "up to date"

# Sort by different fields
gwq status --sort activity
gwq status --sort modified

# CSV output for data analysis
gwq status --csv

# Show status for all worktrees in base directory
gwq status --global

This command is particularly useful for:

• Multi-AI Agent Monitoring: See which worktrees have active changes from different AI agents
• Quick Overview: Instantly understand the state of all your development branches
• Cleanup Identification: Find inactive or stale worktrees that can be removed
• Integration: Use JSON/CSV output for integration with other tools

gwq prune

Clean up deleted worktree information

gwq prune

gwq config

Manage configuration

# Show configuration
gwq config list

# Set worktree base directory
gwq config set worktree.basedir ~/worktrees

# Set naming template
gwq config set naming.template "{{.Repository}}-{{.Branch}}"

gwq tmux

Manage tmux sessions for long-running processes

# List active tmux sessions
gwq tmux list
# Output:
# SESSION            DURATION   WORKING_DIR
# run/npm-myapp      2 hours    ~/ghq/github.com/d-kuro/gwq
# build/make-project 30 mins    ~/projects/myapp
# test/go-tests      1 hour     ~/worktrees/myapp-feature

# Run command in new tmux session
gwq tmux run "npm run dev"

# Run with custom identifier
gwq tmux run --id dev-server "npm run dev"

# Run in specific worktree
gwq tmux run -w feature/auth "make test"

# Run with auto-cleanup (session deleted when command completes)
gwq tmux run --auto-cleanup "make build"

# Attach to running session
gwq tmux attach dev-server

# Kill session
gwq tmux kill dev-server

# Kill all sessions (with confirmation)
gwq tmux kill --all

# Watch session status in real-time
gwq tmux list --watch

# JSON output for automation
gwq tmux list --json

This feature is particularly useful for:

• Development Servers: Keep dev servers running across terminal sessions
• Long Builds: Monitor build processes without keeping terminal open
• Test Suites: Run extensive test suites in the background
• AI Agent Tasks: Let AI agents run long tasks without blocking your terminal
• Remote Work: Sessions persist even if SSH connection drops

gwq task

Manage Claude Code tasks and automated development

# Add a new Claude task with new branch/worktree
gwq task add claude -b feature/auth "Authentication system" -p 75

# Add a task using existing worktree
gwq task add claude -w existing-feature "Continue development" -p 60

# List all tasks
gwq task list

# View task-specific execution logs
gwq task logs                           # Interactive task log selection
gwq task logs exec-a1b2c3               # Show logs for specific execution
gwq task logs --status running          # Filter task logs by status
gwq task logs --date 2024-01-15         # Filter by date

# Worker management
gwq task worker start --parallel 2
gwq task worker status
gwq task worker stop

# View task details
gwq task show task-id

This feature enables:

• Structured Task Management: Create and manage Claude Code tasks with dependency resolution
• Priority-based Scheduling: Organize tasks by priority and dependencies
• Resource Management: Control parallel execution and resource allocation
• Comprehensive Logging: All task executions are logged in JSON format
• Session Management: Persistent tmux sessions for long-running tasks
• Interactive Monitoring: Browse and review task execution history

gwq version

Display version information

# Show detailed version information
gwq version

# Show brief version
gwq --version

Shell Integration

Tab Completion

gwq provides tab completion for all commands, making it easy to discover branches, worktrees, and configuration options.

Setup

Bash:

# Add to ~/.bashrc
source <(gwq completion bash)

Zsh:

# Add to ~/.zshrc
source <(gwq completion zsh)

Fish:

# Save completion script
gwq completion fish > ~/.config/fish/completions/gwq.fish

PowerShell:

# Add to your PowerShell profile
gwq completion powershell | Out-String | Invoke-Expression

After setting up, you can use tab completion:

gwq add <TAB>          # Shows available branches
gwq get <TAB>          # Shows available worktrees
gwq remove <TAB>       # Shows branches and worktrees
gwq config set <TAB>   # Shows configuration keys

Configuration

Configuration file location: ~/.config/gwq/config.toml

[worktree]
# Base directory for creating worktrees
basedir = "~/worktrees"
# Automatically create directories
auto_mkdir = true

[finder]
# Enable preview window
preview = true
# Preview window size
preview_size = 3

[naming]
# Directory name template
# Available variables: Host, Owner, Repository, Branch, Hash
template = "{{.Host}}/{{.Owner}}/{{.Repository}}/{{.Branch}}"
# Invalid character replacement (applied to branch names)
sanitize_chars = { "/" = "-", ":" = "-" }

[ui]
# Icon display
icons = true
# Display home directory as ~ in paths
tilde_home = true

[tmux]
# Enable tmux integration
enabled = true
# Tmux command to use
tmux_command = "tmux"
# History limit for tmux sessions
history_limit = 50000

[claude]
# Claude Code executable
executable = "claude"
# Timeout for Claude execution
timeout = "30m"
# Maximum parallel Claude executions
max_parallel = 3
# Configuration directory
config_dir = "~/.config/gwq/claude"

[claude.task]
# Task queue configuration
queue_dir = "~/.config/gwq/claude/queue"
# Log retention in days
log_retention_days = 30
# Maximum log size in MB
max_log_size_mb = 100
# Auto cleanup old logs
auto_cleanup = true

Advanced Usage

Multiple AI Agent Workflow

# Create multiple worktrees for parallel development
gwq add -b feature/auth
gwq add -b feature/api
gwq add -b bugfix/login

# Launch AI agents in parallel (example with Claude Code)
# Method 1: Using gwq get
# Terminal 1
cd $(gwq get auth) && claude

# Terminal 2
cd $(gwq get api) && claude

# Terminal 3
cd $(gwq get login) && claude

# Method 2: Using gwq exec --stay
# Terminal 1
gwq exec --stay auth -- claude

# Terminal 2
gwq exec --stay api -- claude

# Terminal 3
gwq exec --stay login -- claude

# Method 3: Using direct cd with gwq get
# Terminal 1
cd $(gwq get auth) && claude

# Terminal 2
cd $(gwq get api) && claude

# Terminal 3
cd $(gwq get login) && claude

# Monitor all AI agent activity
gwq status --watch  # Shows real-time status of all worktrees

# Run long-running AI tasks in tmux sessions
gwq tmux run --id ai-migration-auth "claude migrate auth module"
gwq tmux run --id ai-migration-api "claude migrate api module"

# Monitor AI task progress
gwq tmux list --watch

# Attach to see AI output
gwq tmux attach ai-migration-auth

# Method 4: Using gwq task for structured development
# Create structured tasks for automated development
gwq task add claude -b feature/auth "Implement authentication system" -p 80
gwq task add claude -b feature/tests "Create comprehensive test suite" \
  --verify "make test" --verify "make coverage"

# Start worker to process tasks
gwq task worker start --parallel 2

# Monitor task executions
gwq task logs --status running  # See active task sessions

# Review completed task work
gwq task logs  # Interactive selection of past executions

Batch Operations

# List all feature branches from global worktrees
gwq list -g --json | jq '.[] | select(.branch | contains("feature"))'

# Clean up old feature worktrees
gwq list -g --json | \
  jq -r '.[] | select(.branch | contains("feature/old-")) | .branch' | \
  xargs -I {} gwq remove -g {}

# Find worktrees with uncommitted changes
gwq status --json | jq '.worktrees[] | select(.status == "changed") | {branch, changes: .git_status}'

# Export worktree status to CSV for reporting
gwq status --csv > worktree-status-$(date +%Y%m%d).csv

Integration with Git Workflows

# Create worktree for PR review
gwq add -b pr-123-review origin/pull/123/head

# Create worktree for hotfix
gwq add -b hotfix/critical-bug origin/main

# Switch between worktrees quickly
cd $(gwq get)  # Use fuzzy finder to select

Version Information

# Show version information
gwq version

# Show brief version
gwq --version

Directory Structure

gwq organizes worktrees using a URL-based hierarchy similar to ghq, ensuring no naming conflicts:

~/worktrees/
├── github.com/
│   ├── user1/
│   │   └── myapp/
│   │       ├── main/           # Main branch
│   │       ├── feature-auth/   # Authentication feature
│   │       └── feature-api/    # API development
│   └── user2/
│       └── myapp/              # Same repo name, different user
│           ├── main/
│           └── develop/
├── gitlab.com/
│   └── company/
│       └── project/
│           └── feature-x/
└── code.google.com/
    └── p/
        └── vim/
            └── main/

This structure:

• Prevents conflicts: Same repository names from different sources don't collide
• Preserves context: You always know which repository a worktree belongs to
• Scales naturally: Works with any number of git hosting services
• Follows conventions: Similar to how ghq manages repository clones

Requirements

• Git 2.5+ (for worktree support)
• Go 1.24+ (for building from source)
• Terminal with Unicode support (for fuzzy finder)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

Apache License 2.0 - see LICENSE file for details.