Show GN: Multi-agents now dance the talchum

A structure where a main agent calls subagents tends to solidify into subordination rather than collaboration.

Even once roles are divided, subagents end up being treated like internal functions rather than independent participants, making them hard to reuse in other flows. It also becomes increasingly difficult to keep each agent as an entity with its own distinct MCP, skill, system prompt, and model.

To solve this problem, dance-of-tal treats multi-agents as a layer for packaging and reassembling for reuse.

• Tal: the agent's identity and personality
• Dance: reusable capabilities and skills
• Performer: an execution unit combining Tal + Dance + model/tool/MCP/runtime
• Act: the relationships and collaboration flow among multiple Performers

The key point is to treat subagents not as targets for one-off calls, but as components that can be plugged back into different flows for reuse.

Built on top of that, DOT Studio is a Figma-style editor and runtime.

• You can place agents with different configurations on a canvas
• You can design relationships by connecting Tal, Dance, Performer, and Act
• You can visually edit collaboration flows
• You can reduce repetitive work with the Studio Assistant
• You can handle agent configurations running on OpenCode directly in the browser

Ultimately, this project is an attempt to treat multi-agents not as simple orchestration, but as a choreography in which entities with roles and relationships move together.