Dance of Tal (DOT)
The Package Manager for AI Agents Build, share, and choreograph multi-agent systems like you do with
npm.
You're juggling scattered prompts, copy-pasting markdown skills, and manually wiring up MCPs for every new project. It's a mess. We built the package manager for agents. Tested the choreography. Kept what actually scales.
If you're new here, this page gives you the big picture. Once you're ready, jump to Getting Started to set up in about 5 minutes.
š¦ The Package Manager (Not Another Framework)#
We're past the era of monolithic agent frameworks. Just compose blocks like you write package.json.
| Concept | What It Does |
|---|---|
| Tal | The core identity/persona. Who the agent is. |
| Dance | The skill package. Isolated, reusable capabilities pulled straight from GitHub repos (via SKILL.md). |
| Performer | Your composed agent instance. A Tal + arbitrary Dances + a specific LLM model config. |
| Act | The choreography script. Define execution topologies where performers communicate. |
The result? A setup you can explain, version, install, and reuse across machines and teammates. You touch nothing.
Why Teams Use DOT#
Most teams hit the same problems once AI becomes part of everyday work. dot fixes them.
- Prompts become huge and impossible to debug: Break them into Tal (identity) and Dances (skills).
- Team conventions live in random docs: You're actually manually copying prompts across repositories? Wild. Publish reusable assets by URN and share them instantly.
- Setup works in one editor but not another: Portable assets install securely via
dot addanywhere. - Nobody can tell what behavior is actually active: Inspectable workspace and strict contracts.
The Four User-Facing Surfaces#
1. dot CLI#
The core. Use the CLI when you want a fast, file-based workflow. Initialize workspaces, install registry assets, pull skills with dot add, keep them up to date with dot check / dot update, and publish to your namespace.
ā CLI Reference
2. DOT Studio#
Use Studio when you want a visual workspace. Compose performers, build collaborative Acts, and run everything through OpenCode in direct or safe mode. ā Studio Overview
3. Public Registry#
Like npm registry, but for agents. Use the registry to discover and share official and community assets.
ā Explore Registry
Workspace Structure (Local vs Global)#
When you run dot init, DOT creates a local workspace:
your-project/
āāā .dance-of-tal/
āāā assets/ ā Installed & created assets (Tal, Dance, etc.)
āāā drafts/ ā Work-in-progress content
~/.dance-of-tal/
āāā auth.json ā Login token (stored separately)
Recommended Reading Order#
| # | Page | What You'll Learn |
|---|---|---|
| 1 | Getting Started | Install and set up in 5 minutes |
| 2 | Core Concepts | Understand Tal, Dance, Performer, Act |
| 3 | Studio Overview | Visual workspace for building AI systems |
| 4 | Performer Composer | Build performers visually |
| 5 | Act Editor & Runtime | Multi-performer collaboration |
| 6 | CLI Reference | All CLI commands |
| 7 | Publishing Assets | Share your assets with the world |
The Short Version#
DOT is built around a simple idea:
- AI behavior should be portable
- Setup should be inspectable
- Collaboration should be explicit
- Teams should share reusable behavior, not giant prompt blobs
Install dot. Type dot add. Done.