DreamNet Quorum Lab

DreamNet Quorum Lab is a decision-support surface for human-backed AI agents. A user submits a scenario, market, risk question, or workflow decision, and the system convenes a 31-agent quorum. The agents return votes, confidence, disagreement, and reasoning, but they cannot execute trades, move funds, or take external action. If the quorum does not reach the required 28-of-31 threshold, the system fails closed and returns a no-action receipt.

The core idea is controlled autonomy: agents can deliberate, compare evidence, and produce a structured recommendation, but humans remain in charge of execution. Each result becomes an auditable receipt showing the scenario, vote split, safety gates, identity context, and final status. This makes agent output inspectable instead of opaque, and useful for markets, protocol risk, real-world asset verification, product decisions, or any workflow where autonomous agents need governance.

DreamNet Quorum Lab is built as a clean-room Svelte + Vite frontend with a Cloudflare Worker API. The frontend lets users enter a decision scenario and inspect the resulting quorum receipt. The Worker exposes health, forecast, ENS, World, Walrus, and blocked-execution endpoints.

The forecast endpoint simulates a 31-agent quorum and returns a structured paper-mode receipt. The app intentionally blocks execution routes, returning 403 for /api/execute/*, so the demo proves the safety boundary as well as the agent workflow.

Partner-facing integrations are modeled around ENS for readable agent identity and discovery, World for human-backed approval and proof validation, and Walrus for durable receipt storage. The system is designed to show how agent swarms can become safer and more useful when they produce evidence, vote splits, identity context, and receipts before any human-approved action.