wtx
Enables AI coding agents to manage Git worktrees by creating, listing, removing, and cleaning isolated workspaces through MCP tools.
README
wtx
Git worktree manager with a built-in MCP server. Create, inspect, and clean isolated workspaces from your terminal — or let your AI coding agents do it themselves.
Parallel AI coding agents made git stash-juggling obsolete — but raw git worktree add ../foo -b foo && cd ../foo && cp ../.env . is a chore nobody wants to repeat five times a day. wtx turns worktrees into a one-word workflow for humans and for agents:
$ wtx new fix-login
✔ created worktree fix-login
path: ~/code/myapp/.worktrees/fix-login
branch: wt/fix-login (new)
copied: .env, .env.local
Your agent gets the same power through MCP:
{ "name": "create_worktree", "arguments": { "name": "fix-login", "setup": "npm install" } }
Why wtx
- One tool, two interfaces. A fast CLI for you, an MCP server (
wtx mcp) for Claude Code, Cursor, Codex, or any MCP-capable agent. - Zero dependencies. Single Node.js binary that only shells out to
git. Nothing to audit, nothing to break. - Hygiene-aware. Every worktree shows its disk footprint and last activity. Reclaim gigabytes of stale
node_modulesclones withwtx clean --dry-run. - Safe by default. Refuses to delete dirty worktrees, never touches the main checkout, deletes branches only when fully merged, and hides
.worktrees/via.git/info/exclude(no.gitignorepollution). - Batteries included. Copies env files into new worktrees and runs your project's setup hook automatically.
Install
npm install -g wtx # or: brew install wtx (planned)
Requires Node.js ≥ 18 and any modern git.
Quickstart
$ cd myapp
$ wtx new feat-search # isolated workspace at .worktrees/feat-search on branch wt/feat-search
$ cd $(wtx path feat-search) # jump in
$ wtx ls # see every workspace, its state and size
NAME BRANCH STATUS AGE SIZE PATH
myapp* main ✓ clean 2d ago 412 MB ~/code/myapp
feat-api wt/feat-api ● dirty (1) 3h ago 38 MB ~/code/myapp/.worktrees/feat-api
$ wtx clean --dry-run # preview reclaimable space
· feat-api (branch merged, 38 MB)
would free: 38 MB · kept 0 worktree(s)
Commands
| Command | Description |
|---|---|
wtx new <name> |
Create .worktrees/<name> on branch wt/<name>, copy env files, run setup hook |
wtx ls |
Table of worktrees: branch, ahead/behind, dirty state, last-commit age, disk usage |
wtx rm <name> [--force] |
Remove a worktree; refuses if dirty; deletes merged branches |
wtx clean [--dry-run] [--stale-days <n>] |
Remove worktrees whose branches are fully merged and/or inactive |
wtx open [name] |
Spawn $SHELL inside a worktree (--print to just echo the path) |
wtx path [name] |
Print a worktree's absolute path (shell-script friendly) |
wtx mcp |
Run as an MCP server over stdio |
Global flags: --json (machine-readable output), --base <ref>, --setup <cmd>, --no-copy, --no-setup.
Use it from AI agents (MCP)
Add to your MCP client config:
// e.g. claude_desktop_config.json / opencode.json / .mcp.json
{
"mcpServers": {
"wtx": { "command": "npx", "args": ["-y", "wtx", "mcp"] }
}
}
The server exposes four tools: list_worktrees, create_worktree, remove_worktree, clean_worktrees. All accept an optional repo_path; otherwise the server's working directory is used. Responses are JSON, so agents can chain them: create → work → remove without ever switching branches in your main checkout.
Example agent prompt: "Set up a workspace called oauth-fix based on develop, run pnpm install in it, then tell me the path."
Configuration
Drop a wtx.config.json (or .wtxrc) in your repo root:
{
"dir": ".worktrees",
"copyFiles": [".env", ".env.local"],
"setup": "pnpm install",
"branchPrefix": "wt/"
}
| Key | Default | Meaning |
|---|---|---|
dir |
.worktrees |
Where worktrees live, relative to repo root (auto-added to .git/info/exclude) |
copyFiles |
[".env", ".env.local"] |
Files copied from the main checkout into each new worktree |
setup |
(none) | Shell command run inside a new worktree after creation |
branchPrefix |
wt/ |
Prefix for branches created by wtx new |
Safety model
- The main worktree is never removed by
rmorclean. - Dirty worktrees are always preserved by
clean, andrmrequires--force. - Branches are deleted only when fully merged into the default branch (
origin/HEAD→main→masterfallback). - Worktree directories live under
.worktrees/, excluded via.git/info/excludeso your.gitignorestays untouched. - Names are sanitized; traversal like
../evilis rejected.
How it compares
raw git worktree |
TUI managers | wtx | |
|---|---|---|---|
| One-command create + env copy + setup hook | ✗ | partial | ✓ |
| Disk usage & staleness per worktree | ✗ | rare | ✓ |
Machine-readable --json everywhere |
✗ | ✗ | ✓ |
| MCP server for coding agents | ✗ | ✗ | ✓ |
| Zero runtime dependencies | n/a | ✗ (usually heavy) | ✓ |
Development
git clone https://github.com/willkhinz/wtx && cd wtx
npm install
npm test # build + 32 integration/unit tests
node dist/src/cli.js help
Architecture: src/core.ts (engine), src/cli.ts (human UI), src/mcp.ts (agent UI). The MCP implementation is dependency-free newline-delimited JSON-RPC 2.0.
Roadmap
- [ ]
wtx exec <name> -- <cmd>run commands across all worktrees - [ ] tmux/iTerm session spawning per worktree
- [ ] Homebrew formula
- [ ] GitHub Action: comment PRs with per-worktree previews
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.