MINIMAL PRIMITIVES.
NO MORE LANGGRAPH.

If human society can organize billions of people through simple IM interfaces like WeChat, AI Agents should not be locked in complex graph structures (DAGs).

RUN SIMULATION

“ Traditional multi-agent frameworks often force developers to pre-define every node and connection in a static workflow. We believe: Organization should be emergent, not prescribed. ”

01. The Primitives

All Multi-Agent systems can be expressed through two primitives.

We discard tedious state machine definitions and return to the purest communication logic.

Through create(), you can instantly hire or clone a new Agent and get its unique agent_id.

Through send(), you can send asynchronous messages to any known ID. That is all an Agent is.

// Core API Signatures

function create(role: string): Promise<agent_id>;

function send(to: agent_id, msg: string): Promise<void>;

// Usage Example

const

coder_id = await create("coder");
await send(coder_id, "Implement the core loop");

02. Fluid Topology

Built autonomously by Agents, not preset by humans.

Traditional workflows are rigid blueprints. In Agent Wechat, topology 'flows' during execution. When an Agent finds a task too complex, it autonomously hires subordinates. It is a liquid organization, not a stiff gear.

Static Workflow (Old)

Predefined & Brittle

Liquid Topology (Agent Wechat)

Self-evolving & Resilient

03. Direct Collaboration

Intervene at any level, just like chatting.

Traditional Agent systems are black boxes. In Agent Wechat, humans have a global perspective. You can start a conversation with any sub-agent directly via IM. This flat intervention makes complex topologies observable, debuggable, and actionable.

Human Collaboration Term

Wait! Make it 3D instead.

Ready for the era of Agent Social?

Explore Interactive Demo