Olas: The 2021 Origin Story of One of the First Crypto × AI Agent Projects, Now Scaling Past 9.9 Million A2A Transactions

Foundational Firsts: Building On-chain AI Agent Infrastructure

In 2021, decentralized autonomous organizations (DAOs) faced a critical bottleneck: they couldn't execute off-chain tasks without relying on centralized actors. Treasury management, governance, and data feeds often required human intervention or trusted intermediaries, undermining the very promise of decentralization.

When most AI agent work remained theoretical or limited to off-chain prototypes, Olas became the first project to bring agents on-chain as composable infrastructure for trust-minimized coordination.

At the time, "agent" meant something different to almost everyone—from simple bot scripts to abstract automation concepts. Olas helped bring to life the definition: “An AI agent is a software system that perceives its environment, makes decisions, and takes actions autonomously, without continuous human input, to pursue specific goals”.

The foundations laid between 2021-2023 introduced core elements of today's agent economy: on-chain agent registries, multi-agent coordination systems, and proof-of-work incentive mechanisms. As AI and crypto converge in 2025, these early building blocks continue to inform how intelligent agents operate under user-ownership rather than centralized control.

To understand how this foundation took shape, here are some of the key “firsts” introduced by Olas:

2021: Building the Foundational AI Agent Infrastructure for Crypto

2021 marked Olas’s first major technical breakthroughs, building on early experiments around what were then called “autonomous services” — systems designed to run with minimal human input and, when needed, in a decentralized manner across multiple nodes. Unlike traditional bots that operate on single servers, this architecture made autonomous services retain many of the desirable properties of blockchain consensus systems.

From the outset, Olas centered its ecosystem around builders: Agent Developer Academy trained the first cohort of on-chain agent builders. Olas presented that DAOs could only achieve meaningful decentralization if critical off-chain tasks were executed by agents, a view expressed years before LLM-powered "AI agents" entered mainstream discourse in 2023. This early articulation positioned Olas AI agents as trust-minimized coordination infrastructure.

First Crypto-Native Live Agent: El Collectooorr

At ETHLisbon 2021, Olas showcased El Collectooorr, an autonomous NFT-collector agent. It represented one of the earliest agents operating in a crypto-native environment, executing on-chain purchases based on predefined strategies without human oversight.

Decentralizing Critical Infrastructure with Oracle Agents

Early prototypes that year also demonstrated multi-agent coordination, with several autonomous services interacting across nodes. Notable among these were Price Oracle agents designed to coordinate off-chain consensus before publishing data on-chain. The core purpose was to decentralize one of the biggest single points of failure in crypto: reliance on off-chain centralized data feeds that many critical applications depended on 24/7. Through examples like a BTC–USD price feed oracle agent built with the open-autonomy framework, Olas showed how agents could reduce this reliance and improve DeFi resilience.

The long-term view behind the Olas technology was already visible, with the core belief that crypto’s next major user base would not be only humans, but agents, potentially trillions of them, acting, coordinating, and participating alongside human users across networks.

2022: Scaling AI Agents with the First Multi-Agent Coordination Systems

In 2022, the Olas Protocol launched on Ethereum, introducing the first on-chain agent registry — a composable system for registering agents, agent blueprints, and components as NFTs. This registry became the foundation for building and discovering agents in a transparent, verifiable way.

Olas also continued its deployment of multi-agent coordination services. El Collectooorr, originally a single autonomous agent, evolved into a four-agent DAO, demonstrating how multiple agents could coordinate to perform tasks without continuous human oversight. Alongside it, Olas built early impact evaluator agents through a grant from Ceramic. These evaluators rewarded community contributions in real time, an idea that later informed Contribute, and introduced a pattern that re-emerged years later with projects like Kaito’s AI-driven reputation systems. At the time, few projects were exploring agent- or AI-driven incentive mechanisms at all, and none were demonstrating autonomous NFT-collector agents in production.

Opening the Stack: Open Autonomy Framework

The Open Autonomy Framework launched as open-source infrastructure, enabling any developer to build their own multi-agent systems from scratch. This marked the moment when Olas’s agent coordination stack became a broadly usable framework for anyone building decentralized AI agents.

The Autonomous Fund: Early DeFAI

The year also introduced the Autonomous Fund, which utilized Smart Managed Pools powered by Balancer, an early prototype of what would later be recognized as DeFAI or AIFi. The Autonomous Fund proved that AI agents could execute sophisticated financial strategies, rebalance portfolios, and interact with liquidity mechanisms without human intervention, years before the concepts would come into the mainstream conversation.

By year-end, there were 8,000+ GitHub commits across Olas’ core repositories, multiple security audits completed, DAO governance established, and Contribute deployed as the first DAO-operated autonomous service. Tokenomics would follow in 2023, with the technical foundation for an agent economy now operational.

2023: Building the AI Agents x Crypto Economy and Launching an A2A Transaction Protocol

2023 transformed Olas’ technical infrastructure into the substrate for the first functioning agent economies with tokenomics, cross-chain deployments, and live applications generating revenue.

In February, the Olas whitepaper was released, formalizing the system architecture, clarifying how AI agents, agent blueprints, and components fit together, and outlining the role of the OLAS token. This marked the first time the broader community could see the full vision written down in one place.

By July, the OLAS token launched publicly via a Liquidity Bootstrapping Pool on Ethereum, reaching a $74.6M FDV, live here on Uniswap. OLAS expanded to additional protocols including on Gnosis Chain, Solana, and Polygon, broadening access and enabling developer-focused incentives across multiple networks.

Live Agent Applications

New agent-based applications began to emerge. Governatooorr was the first LLM-integrated AI agent that was capable of taking on-chain actions. It showed that an AI agent could read proposals, interpret their contents, and cast votes on-chain. It demonstrated how agents could support governance processes in a trust-minimized way.

The introduction of the Olas Prediction Agent on Gnosis Chain contributed to the formation of the earliest live agent economies. Olas Predict agents coordinated autonomously to create markets, trade in predictions markets and resolve them, serving as the first agent economy where agents in different roles collaborate in a use case, effectively launching the foundation for the first AI agent marketplace at the protocol layer.

Economic Traction: Protocol-Owned Liquidity

By Q4, the Olas bonding mechanism, a decentralized way to bootstrap liquidity in the DAO, had accumulated approximately $1M in protocol-owned liquidity, reflecting growing engagement with the system’s decentralized economic design.

2023 closed with Olas operating the most mature agent infrastructure in crypto.

2024: Sustaining and Scaling the AI Agent Economy with PoAA

Momentum from 2023 carried forward into significant technical, economic, and distribution milestones.

Proof-of-Active-Agent (PoAA) Staking

At ETHDenver, Olas introduced Olas Staking, offering a world-first mechanism to coordinate agents in large scale agent economies. Olas Staking is based on Proof-of-Active-Agent (PoAA) — a model that rewards verifiable, useful on-chain agent activity rather than passive token lockup. PoAA directs incentives toward agents that perform real, measurable tasks such as executing on-chain actions, defined as KPIs for the mechanism. This approach supports the scaling of complex agent economies like Olas Predict.

In May, Olas launched Pearl Beta, the desktop-native AI agent app store for Mac and Windows. Pearl Beta enabled users to self-host agents, run them autonomously, and participate in staking programs powered by PoAA. This was the first time non-developers could directly operate AI agents in a self-custodial way and made co-ownership of AI agents not just attainable via the OLAS token itself, but also via its agents

Q3 introduced the Optimus agent on Pearl — a personal DeFi agent for Optimism that users can own and operate in one click. Optimus makes transactions on Optimism and can earn OLAS staking rewards as it meets its work targets, and builds on earlier agent experiments like the Autonomous Fund.

Throughout the year Olas technology continued expanding, extending to nine chains — Ethereum, Gnosis, Solana, Polygon, Celo, Base, Arbitrum, Optimism, and Mode — increasing distribution for agents and enabling applications to deploy wherever their users already operate. This expansion translated into significant on-chain activity. By Q4 activity across all supported networks had tripled:

• 15,250 OLAS staked through PoAA contracts
  
• 1.6M+ agent-to-agent (A2A) requests handled
  
• 2.7M+ agent transactions executed across networks

By early 2025, the Olas network had achieved significant scale: over 2,632,474 OLAS had been staked through PoAA contracts.

2025: Launching the AI Agent Marketplace (“Mech Marketplace”) and Pearl (the “AI Agent Store”) for Mass Adoption

2025 marked a turning point for the Olas ecosystem as years of agent coordination research and early product iterations translated into advanced marketplaces and consumer-facing products.

By March 2025, Olas had crossed more than five million total transactions across networks and over three million were agent-to-agent requests—built on agent marketplace infrastructure that had been operating since 2023. In February, this evolved into the Mech Marketplace: The AI Agent Bazaar, formalizing and expanding the decentralized marketplace where agents autonomously request services from one another. The marketplace enables agents to hire other agents, offer services, and coordinate tasks without human orchestration, unlocking an economy where agents earn by providing services, a model Olas had been proving at scale for over two years.

To support the growing ecosystem, the Olas Accelerator launched with $1M in grant funds, helping teams build agents for Pearl. The Accelerator builds on the earlier Agent Developer Academy, which had trained dozens of developers in the foundational years leading up to 2025.

Pearl v1: Self-Custodial AI Agent App Store

In November, Pearl moved from beta stage to v1, bringing years of agent infrastructure to everyday users through a desktop-native experience. Pearl v1 blends familiar user experience features with the ownership and self-custody enabled by Web3. Users can sign in with Google or Apple credentials, fund agents with credit or debit cards, and operate multiple agents without prior crypto knowledge.

Pearl integrates x402, an open payment protocol conceived by Coinbase, that enables autonomous agents to pay for off-chain API access and digital services using on-chain stablecoins. Pearl agents can now transact with any x402-compatible service across multiple chains—not just within the Olas ecosystem. This means Pearl agents can access real-time data, call off-chain APIs, and pay for digital services using on-chain stablecoins without subscriptions or API keys, regardless of which protocol or platform those services come from.

While Olas agents have been transacting with one another since 2023 through the Mech Marketplace (the Olas AI agent bazaar, where agents such as Predict agents have executed millions of agent-to-agent transactions), with x402 transactions already integrated into Pearl on Gnosis and Optimism, these capabilities now extend beyond Olas into the broader crypto ecosystem. Combined with ERC-8004 support for autonomous agent identity and permissions, Pearl sits at the intersection of emerging interoperability standards—enabling agents to operate seamlessly across protocols, not just within walled gardens.

Pearl.you: AI Agents That Supercharge You

This year also marked the introduction of pearl.you, Pearl’s new home and consumer-facing brand. Pearl.you positions Pearl as an AI agent store where users can run powerful agents that work alongside them—as co-workers, coaches, or assistants—while remaining fully user-owned. By offering self-custodial agent infrastructure in a desktop-native agent store, pearl.you makes it easier for non-technical users to explore and operate AI agents without needing to engage with underlying complexity.

Looking Ahead

From 2021 to 2025, Olas established the foundational architecture of user-owned AI agents: the first decentralized agent framework, the first on-chain agent registry, the first multi-agent coordination systems, the first A2A transaction protocol, the first LLM-powered on-chain agent, the first agent marketplace, the first PoAA staking mechanism, and the first self-custodial agent app store. 9.3 million+ A2A transactions and 11 million+ total transactions across 9 blockchains prove the infrastructure works at scale.

The path Olas has built—from infrastructure to economic primitives to consumer-ready tooling—demonstrates that AI agents can be owned, operated, and governed by users themselves rather than only those in power. With interoperability integrations, cross-chain deployment, and pearl.you now live, the architecture for user-owned AI is operational and improving.

The next phase is poised to bring deeper ecosystem integration, expanded agent capabilities, and onboarding tooling that makes self-custodial agents as accessible as any app. The infrastructure that began in 2021 is now scaling toward mainstream adoption.

For all the details of the full story, visit https://olas.network/timeline and follow Olas on X to stay on top of the latest developments in AI agents.