> For AI agents: the complete documentation index is available at /llms.txt, the full documentation bundle is available at /llms-full.txt.

# Tack: One setup. Every coding agent.

Write a skill once. Tack links it wherever your agents look, in every project.

## See it run

A coding agent sets Tack up from one prompt, on a machine that lives in this page.

In the landing page's terminal, the same setup walks through these commands:

```sh
$ tree ~/skills
$ tack init
$ tack source add ~/skills --skills .
$ tack source add ~/skills --skills . --apply
$ tack source add ~/team-docs --docs handbook --commands commands --apply
$ cd ~/code/app && tack project add app
$ cd ~/code/api && tack project add api
$ tack sync --all
$ tack sync --all --apply
$ tree ~/code/app
$ cd ~/code/app && tack docs attach handbook
$ tack sync --project app --apply
$ cd ~/skills && git pull
$ tack sync --all --apply
$ ls -l ~/code/api/.claude/skills
```

## One copy, linked everywhere.

Edit a skill once. Every agent and project that links it sees the change. Teammates get it with a pull.

### Every agent, every project, every worktree.

One skill folder, linked into Claude Code, Codex, Cursor, Copilot CLI, Gemini CLI, Cline, OpenCode and Pi: for you, and in each project you register, worktrees included. Turn on a new agent and one sync links every skill you use. Turn it off and only Tack's links go. Each agent also gets the `use-tack` skill, so it can run Tack for you.

```sh
tack adapter enable codex
```

### Share skills like code.

Keep the team's skills and docs in a Git repository. Each teammate clones it once and links it into the agent they use. Edits arrive with `git pull`. New skills wait for your sync preview. Each project picks its own skills, and every worktree of a project gets the same links.

```sh
tack source add ~/code/acme-agent-skills
```

### Docs your agents can find.

Attach a docs folder to a project and every worktree gets it under `.tack/docs`. Agents search it with `tack docs search`. You read it in a local reader, and a passage you copy keeps its file and line numbers.

```sh
tack docs attach handbook --project app
```

### Also in Tack

- **Scripts, run by name.** Each run gets only the secrets it declares and leaves a history. `tack script run report --project app`
- **Skills from GitHub.** Copies a skill into your source and records where it came from. `tack import add github:org/repo --skill pdf --into team-skills`
- **One file for CI.** Packs a reviewed set of skills into one bundle that sets up a fresh machine without fetching. `tack profile export`

## Tack is the gear, not the horse.

Your coding agent does the running. Tack fits your skills onto whichever you ride, and takes them off clean.

- **Write it once.** Skills, commands and docs live in folders you own, or in a Git repository your team shares. `~/skills/review/SKILL.md`
- **Fit every agent.** Tack links your skills into each agent you use. Linked, never copied: edit a skill once and every agent sees it. `.claude/skills/review → ~/skills/review`
- **Stay on track.** Every sync shows its plan before it writes, and keeps each project and its worktrees current. Tack removes only the links it made. `tack sync --all --apply`

## Get started

Paste this prompt into your coding agent:

> Install Tack with `npm install --global @tackup/cli`, then run `tack guide onboarding` and follow it to set up my machine: ask me which folders hold my skills, commands or docs and which projects to link, explain each step, show me its preview, and wait for my yes before you apply it or run any other command that writes without a preview.

Or run the commands yourself. tack init sets up Tack at once. Every later command that changes links prints a preview first.

```sh
npm install --global @tackup/cli
tack init
tack source add ~/agent-setup --skills skills
tack source add ~/agent-setup --skills skills --apply
tack project add app
tack sync --project app
tack sync --project app --apply
```

Requires Node 22.12 or later, or Bun. Your files are not moved or copied.

## Questions

- **Will it touch my files?** Commands that change links to your files preview first and write only with --apply; tack init writes at once to set up Tack and link its own skills. Tack writes symlinks, never copies, and never overwrites a file it did not create.
- **I can symlink my skills myself.** You can, for one agent and one folder. Tack does it for three agents, at user level and in every project and worktree, keeps a record of what it created, and removes only that.
- **What if I stop using Tack?** Your skills are plain SKILL.md folders in your own directories. tack state reset --apply removes every link Tack created and its state. Your files stay.
- **Does it call a model or send data?** No. Tack is a local CLI. It calls no model and has no telemetry. Only tack import add reaches the network, to fetch from GitHub.
- **Can I share a setup with my team?** Yes. Keep the team's skills and docs in a Git repository. Each teammate clones it, registers it with tack source add, and gets edits with git pull. Fixed profiles and portable bundles reproduce an exact setup.
- **Is Tack open source?** Yes. Tack is open source under the MIT license.

Read the docs: [/docs/](/docs/)
