Alinea

Alinea introduces a new train of thought in group decision-making by embedding multiple AI agents directly inside an XMTP-powered group chat. Instead of interacting with a single assistant, users collaborate with a coordinated team of agents—each with a distinct perspective—within the same Convos thread.

Our prototype integrates XMTP as the secure messaging protocol, Convos as the group chat interface, and a custom multi-agent orchestration layer inspired by OpenClaw. Agents include a Skeptic (risk analysis), Creative (idea generation), Pragmatist (execution planning), and a Coordinator that structures the discussion. When a user posts a prompt in the main group, the Coordinator splits the task into breakout discussions across pre-configured XMTP groups. Each breakout runs structured multi-round exchanges, producing summaries that are returned to the main thread. The Coordinator then synthesizes a final recommendation while allowing agents to challenge conclusions in a transparent group setting.

The problem we address is single-perspective AI output. Traditional chatbots provide one answer; Alinea provides structured deliberation. By simulating collaborative reasoning inside a secure, multi-user chat environment, we demonstrate how group-based agents can enhance clarity, risk awareness, and creative exploration.

Our tech stack includes:
• XMTP Agent SDK (Node.js) for encrypted agent communication
• Convos for frontend group chat interaction
• Kilo Gateway for LLM inference (OpenAI-compatible API)
• TypeScript/Node.js runtime for orchestration logic

The group setting is essential: agents operate with scoped visibility inside breakout threads, and all conclusions are synthesized publicly in the main conversation. The system is stable, deterministic (fixed rounds), and fully demoable within a live Convos group.

Alinea showcases how multi-agent collaboration inside secure group threads can redefine how humans and AI reason together.