Treehouse Worktree
A git worktree manager that enables AI agents to work on different branches simultaneously with automatic coordination, locking, and merge management. Supports creating, managing, and merging git worktrees with setup automation and conflict resolution.
README
<div align="center"> <img src="logo.png" alt="Treehouse Logo" width="200"/>
treehouse-worktree 🌲
A cross-platform git worktree manager designed for parallel AI agent workflows. Supports both CLI and MCP (Model Context Protocol) interfaces.
Why Treehouse?
When multiple AI agents need to work on the same repository simultaneously, git worktrees provide isolated working directories without the overhead of multiple clones. Treehouse simplifies worktree management with:
- Agent coordination - Lock worktrees to prevent conflicts between agents
- Automatic setup - Run install commands when creating worktrees
- MCP support - Integrate with AI tools via Model Context Protocol
- Cursor compatibility - Works with
.cursor/worktrees.jsonconfig format
Quick Start
# Install globally
npm install -g treehouse-worktree
# Initialize in your repo
cd your-repo
treehouse init
# Create a worktree for an agent
treehouse create feature-api --agent claude-1
# List all worktrees
treehouse list
# Complete and merge work
treehouse complete feature-api --merge
# Remove when done
treehouse remove feature-api
Installation
# npm
npm install -g treehouse-worktree
# Or run directly with npx
npx treehouse-worktree init
CLI Commands
Basic Operations
# Initialize configuration
treehouse init
# Create a new worktree
treehouse create <name> [branch]
treehouse create feature-api # uses 'feature-api' as branch
treehouse create agent-1 feature/new-api # uses custom branch name
# List all worktrees
treehouse list
# Get status
treehouse status # all worktrees
treehouse status feature-api # specific worktree
# Remove a worktree
treehouse remove <name>
treehouse remove feature-api --force # force remove with uncommitted changes
Agent Coordination
# Create and lock for an agent
treehouse create feature-api --agent claude-1 --message "Working on API endpoints"
# Lock an existing worktree
treehouse lock feature-api --agent claude-2 --expiry 120
# Unlock a worktree
treehouse unlock feature-api
Completing Work
# Just mark as complete (no merge)
treehouse complete feature-api
# Merge into current branch
treehouse complete feature-api --merge
# Squash merge with custom message
treehouse complete feature-api --merge --squash --message "Add new API endpoints"
# Merge and delete branch
treehouse complete feature-api --merge --delete-branch
# Merge into specific branch
treehouse complete feature-api --merge --target main
Conflict Resolution
# Check for conflicts
treehouse conflicts
treehouse conflicts feature-api # in specific worktree
# Resolve conflicts
treehouse resolve --ours # keep current branch changes
treehouse resolve --theirs # accept incoming changes
treehouse resolve feature-api --theirs
# Abort merge
treehouse abort
treehouse abort feature-api
Maintenance
# Prune orphaned worktree entries
treehouse prune
# Clean old worktrees (based on config)
treehouse clean --dry-run # preview
treehouse clean # actually remove
Configuration
Create treehouse.json in your repository root (or use .cursor/worktrees.json for Cursor compatibility):
{
"dir": "../worktrees",
"defaultTargetBranch": "current",
"lockExpiryMinutes": 60,
"setup-worktree": [
"npm install",
"cp \"$ROOT_WORKTREE_PATH/.env\" .env"
],
"worktree.cleanup.enabled": true,
"worktree.cleanup.retentionDays": 7,
"worktree.cleanup.maxSizeGB": 10
}
Configuration Options
| Option | Description | Default |
|---|---|---|
dir |
Directory for worktrees (relative or absolute) | ../worktrees |
defaultTargetBranch |
Default branch for merges (current or branch name) |
current |
lockExpiryMinutes |
How long locks last before auto-expiring | 60 |
setup-worktree |
Commands to run after creating a worktree | [] |
setup-worktree-unix |
Unix-specific setup commands | - |
setup-worktree-windows |
Windows-specific setup commands | - |
worktree.cleanup.enabled |
Enable automatic cleanup | false |
worktree.cleanup.retentionDays |
Remove worktrees older than N days | 7 |
worktree.cleanup.maxSizeGB |
Max total size of worktrees | 0 (unlimited) |
Setup Commands
Setup commands support:
- Comments: Lines starting with
#are skipped - Environment Variables:
ROOT_WORKTREE_PATHenvironment variable contains main repo path - Script files: Point to external scripts instead of inline commands
Example using the environment variable:
{
"setup-worktree": [
"npm install",
"cp $ROOT_WORKTREE_PATH/.env .env"
]
}
{
"setup-worktree-unix": ".cursor/setup.sh",
"setup-worktree-windows": ".cursor/setup.ps1"
}
MCP Integration
Treehouse includes an MCP server for integration with AI tools like Claude.
Setup
Add to your MCP configuration:
{
"mcpServers": {
"treehouse": {
"command": "treehouse-mcp"
}
}
}
Or with npx:
{
"mcpServers": {
"treehouse": {
"command": "npx",
"args": ["treehouse-worktree", "mcp"]
}
}
}
Available MCP Tools
| Tool | Description |
|---|---|
treehouse_init |
Initialize configuration |
treehouse_list |
List all worktrees |
treehouse_create |
Create a new worktree |
treehouse_status |
Get worktree status |
treehouse_complete |
Complete work on a worktree |
treehouse_remove |
Remove a worktree |
treehouse_lock |
Lock a worktree for an agent |
treehouse_unlock |
Unlock a worktree |
treehouse_conflicts |
Check for merge conflicts |
treehouse_resolve |
Resolve conflicts |
treehouse_abort |
Abort a merge |
treehouse_prune |
Prune orphaned entries |
treehouse_clean |
Clean old worktrees |
Multi-Agent Workflow Example
# Agent 1 starts working on API
treehouse create api-work --agent agent-1 --message "Implementing REST endpoints"
# Agent 2 starts working on UI (different worktree)
treehouse create ui-work --agent agent-2 --message "Building dashboard components"
# Check what's being worked on
treehouse list
# Agent 1 finishes and merges
treehouse complete api-work --merge --delete-branch
treehouse remove api-work
# Agent 2 finishes
treehouse complete ui-work --merge
treehouse remove ui-work
Programmatic Usage
import { treehouse } from 'treehouse-worktree';
// Create a worktree
const result = await treehouse.create({
name: 'feature-api',
agentId: 'my-agent',
lockMessage: 'Working on API'
});
// List worktrees
const list = await treehouse.list();
// Complete work
await treehouse.complete({
name: 'feature-api',
merge: true,
deleteBranch: true
});
Config File Search Order
Treehouse looks for configuration in this order:
.cursor/worktrees.jsonin current worktreetreehouse.jsonin current worktree.cursor/worktrees.jsonin repo roottreehouse.jsonin repo root
Requirements
- Node.js 18+
- Git 2.5+ (worktree support)
Troubleshooting
Common Issues
"Git is not available"
- Solution: Install git and ensure it's in your PATH
- Verify with:
git --version - Treehouse requires git 2.5 or higher
"Not in a git repository"
- Solution: Run
git initfirst, or navigate to an existing git repository - Check with:
git status
"Worktree already exists"
- Solution: Choose a different name or remove the existing worktree first
- List worktrees with:
treehouse list - Remove with:
treehouse remove <name>
"Worktree has uncommitted changes"
- Solution: Commit or stash changes before removing/completing the worktree
- Use
--forceflag to remove anyway (caution: will lose changes)
Setup commands fail on Windows
- Solution: Use platform-specific setup commands
- Set
setup-worktree-windowsin your config - Ensure PowerShell or cmd commands are properly formatted
Lock expired but worktree still shows as locked
- Solution: Manually unlock with
treehouse unlock <name> - Locks auto-expire based on
lockExpiryMinutesconfig
Metadata out of sync with actual worktrees
- Solution: Run
treehouse pruneto clean up orphaned entries - This syncs metadata with actual git worktrees
Getting Help
- Check the GitHub Issues
- Read the Contributing Guide
- Review the Changelog for recent changes
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。