Fresco Agent

An autonomous operations agent that plans, delegates and verifies work toward one long-term goal — with a human always in control of the rules.

What it does

• Keeps a single, unlimited task tree with stable hierarchical IDs (1, 1.1, 1.1.1) and works through it breadth-first, one finite unit of work at a time.
• Assigns each task to a specialist agent — developer, researcher, video producer, fundraiser, communications, watchdog — and records every reassignment.
• Maintains a deep memory tree that refines and corrects knowledge without ever destroying the original evidence.
• Marks work as solved only when the task's own success criteria are verified, not just when its subtasks are done.

How it stays safe

• An independent controller with an explicit permission matrix; the agent cannot change its own permissions.
• Self-improvement only through sandbox, tests, checkpoint and automatic rollback.
• Model-agnostic core: the language model is chosen manually by a human, never by the agent itself.
• Idempotent recovery — a restart never repeats an external action like a message, payment or publication.
• A single HTML control panel showing tasks, memory, agents, questions to the human, logs and permissions.

Why

Fresco Agent is the operational core for Humanity 2.0 — a platform where people propose and vote on ideas that AI agents help bring into the real world. The project is inspired by the resource-based, science-driven thinking of Jacque Fresco.