ZeroicAI

We built a multi-agent framework in Rust. BDI cognition. FIPA messaging. 8 organizational patterns. A supervised async runtime. Meet ZeroicAI

The Holarchy pattern models entities that are simultaneously a whole and part of a larger system. A city is a whole, yet part of a country. A country is part of a continent. Each level is autonomous while remaining subordinate to the next.

New website is now live.

New standard website coming soon. It's 90% ready.

The ExponentialBackoff restart policy doubles retry delays automatically: 1s → 2s → 4s → 8s → 16s... Inspired by TCP retransmission and cloud retry strategies, it prevents repeatedly failing agents from overwhelming the system.

Every AgentId is a UUID v4. The odds of a collision are about 1 in 5.3 × 10³⁶ - effectively zero, even if you created trillions of agents every second for billions of years.

What if AI agents had their own economy? We're not talking about chatbots. We're talking about autonomous agents that discover each other, negotiate prices, sign agreements, exchange payments, and compete for work without human intervention. Agent A needs price data. It finds Agent B. They negotiate. Payment is settled. The task gets done. Agent B starts losing contracts. It lowers its price. Demand returns. The market adjusts itself. Supply. Demand. Competition. Not between humans. Between AI agents. That's what we're building next with the ZeroicAI framework. The agent economy starts here. 🦀

We're working on adding a Solana page to the website that showcases on-chain activity, helping $0ICAI holders better understand ZeroicAI.

UtilityFunction in z-cognition is based on expected utility theory from economics. It scores available choices numerically, and the highest-scoring option becomes the selected action.

ZeroicAI's CircuitBreaker has three states: Closed, Open, and Half-Open. It's the same resilience pattern popularized in distributed systems, adapted for autonomous agents with beliefs, goals, and reasoning loops.

z-messaging uses bounded mailboxes throughout the system. No unbounded queues. No memory blowups when producers outpace consumers. Backpressure is built in from day one, not bolted on after production issues.

Please join our official Tg chat using the link below: Tg Chat: t.me/ZeroicAI

ZeroicAI agents have 5 lifecycle states: Created → Initialized → Running → Paused → Stopped. The type system enforces valid state transitions. You can't call execute() on an uninitialized agent, the compiler prevents it, while the runtime manages the lifecycle.

The "0" in our ticker stands for "Zero."

Just so you know, we started working on ZeroicAI five months ago. Our GitHub activity proves it.

54k views on our pinned tweet. We'll be releasing more demos soon.

We built a Solana trading swarm in Rust. 6 autonomous agents. Live coordination. Zero human intervention. • PriceOracle broadcasts SOL/USDC prices • 3 trading agents use BDI reasoning to decide whether to buy, hold, or sell • A sealed-bid auction settles every trading round • TraderGamma crashes twice. The Supervisor automatically detects, restarts, and reintegrates it. All in ~250 lines of Rust. cargo run --example solana_swarm github.com/ZeroicAI #Solana #Rust #AIAgents

I've come to realize that we may need community channels where $0ICAI holders can communicate and engage with one another. Please vote for your preferred option:

You literally cannot spawn an unsupervised agent in ZeroicAI. Every agent is automatically registered with the Supervisor at spawn. Restart policies are built in: Never, Always, OnFailure, or ExponentialBackoff. Erlang/OTP philosophy. Rust safety.

@GoodJonrd60780 Not our fault. Impatient holders sell.

@ZeroicAI Because it's a scam. It has already gone down the drain. Can't you see that? 🤣🤣

Every agent in ZeroicAI implements Send + Sync. The agent contract is enforced at compile time, preventing data races before your code ever runs. That's Rust working in your favor.