TL;DR: The convergence of Crypto and AI is giving rise to a burgeoning Onchain AI economy, an ecosystem of blockchain apps and services run by autonomous AI agents. And while Decentralized AI projects have seen the lionshare of funding and growth over the last 18 months, we believe Onchain AI is gaining momentum and signaling the next phase of rapid innovation at this intersection. The importance of Onchain AI lies in expanding crypto to potentially billions of AI-powered participants. Every autonomous AI agent is like a new “user” of blockchains that can operate 24/7 and make sophisticated decisions, driving a significant increase in onchain activity and growth. Through investing in Onchain AI, Coinbase Ventures is supporting the builders of this future agent-based economy, ultimately paving the way to a new “Agentic Web”.

Coinbase Ventures portfolio companies are denoted with an asterisk (*) when first referenced in the article below.   

In Oct 2024, Coinbase Ventures published its thesis on the convergence of Crypto x AI, in which we observed that blockchains and AI have complementary strengths - blockchains offer decentralization, censorship resistance, verifiability, and user ownership, while AI brings powerful data processing, reasoning, and automation capabilities. We believe this synergy can transform the way humans and machines interact in the digital economy, ultimately leading to the emergence of an “Agentic Web,” where AI agents operate on top of crypto infrastructure rails to drive significant economic activity and growth. 

A critical distinction from that thesis is between Decentralized AI and Onchain AI. Decentralized AI (“Crypto → AI”) refers to building generic AI infrastructure that inherits the open, peer-to-peer properties of blockchain networks​. This includes efforts to democratize access to compute, data, models, and training so that AI development isn’t monopolized by a few large firms. These DeAI resources are also an enabler of Onchain AI (“AI → Crypto”) – an ecosystem of apps and services that embed AI to power new and existing blockchain use cases (e.g., trading agents, onchain portfolio managers, DeFi abstraction, etc). While Decentralized AI projects have seen the lionshare of funding and growth over the last 18 months, we believe Onchain AI is gaining momentum and signaling the next phase of rapid innovation at the intersection.

An introduction to Onchain AI

Over the past year, we saw an AI agent (i.e., Truth Terminal), equipped with a self-custody wallet, create an internet-native religion and launch a memecoin that surpassed a $950M market cap, making it the first AI agent millionaire. Today, there are roughly ~1.6K+ agents accounting for a combined ~$11B+ market cap, per cookie.fun. Overall, we’re seeing the rapid emergence of AI agents (and associated “agent tokens”) take over social channels, some with utility, and rapidly transform Onchain AI from concept to burgeoning reality.  In particular, three interrelated concepts are gaining prominence: Onchain AI Agents, Onchain AI Applications, and Agentic Commerce.

Onchain AI Agents are autonomous programs (powered by AI models) that can perform onchain actions. Think of an AI agent as a smart software robot with a crypto wallet – it can hold tokens, interact with smart contracts, trade assets, or even vote in a DAO, all based on its programming and goals. Unlike the isolated AI chatbots we often see on our social channels today, these agents can learn, reason, and act within the onchain economy. 

Onchain AI Apps refer to blockchain-based apps that integrate AI into their core functionality. For instance, AI can be embedded into DeFi protocols to optimize yield, into games to control NPC behavior, or into decentralized social networks or consumer apps to hyper-personalize user content. We’ll explore these examples shortly, but the key is that these apps aim to seamlessly blur the line between blockchain and AI-powered logic.

Agentic Commerce is an emerging mode of commerce where AI agents transact with each other (and with humans) using blockchain rails. This is a paradigm shift from manual, search-based transactions to more autonomous, intent-driven and personalized transaction experiences. Agents will be the shoppers, negotiators, and service providers, handling transactions at software speed but aligned with human intent. Crucially, blockchains provide these agents with identities, wallets, stablecoins as a payment currency, and smart contract frameworks to programmatically transact. 

In short, the importance of Onchain AI lies in opening crypto to potentially billions of AI-powered participants. Every autonomous AI agent is like a new “user” of blockchains – one that can operate 24/7 and make sophisticated decisions, setting the stage for significant onchain activity and growth. With that said, let’s take a look at the burgeoning Onchain AI ecosystem and dive into the building blocks (New infrastructure services and types of onchain agents), the emerging onchain AI applications, and how commerce itself may be transformed.

Agents

Onchain AI agents are at the heart of the “Agentic Web.” These are AI-driven entities that can perceive, decide, and act within onchain economies. To understand their rise, let’s break down the enabling infrastructure needed to make onchain agents a reality and then look at the types of agents coming to life.

Agent Infrastructure & Services

Building a capable onchain AI agent is complex – it requires a new set of services and tools that leverage foundational DeAI infrastructure resources (e.g., compute, data, models, intelligence, etc) to enable an open, autonomous agent ecosystem. These services make it easier to create, deploy, discover, and operate autonomous onchain agents by abstracting away complexity and offering reusable components. Below we outline key emerging categories of agent infrastructure and their roles in the onchain AI stack.

• Trusted Execution Environments – To truly operate autonomously and securely, onchain AI agents need environments where their execution is tamperproof, verifiable, and independent of any centralized party. Trusted Execution Environments (TEEs), such as Intel SGX or decentralized alternatives like Eternis*, Fleek*, or Phala Network, offer a hardware-secured enclave where an agent’s code and data can be processed confidentially—even from the agent creators themselves. Agents running inside TEEs are shielded from external interference and can produce cryptographic attestations proving they acted as programmed. As the agentic economy scales, embedding sovereignty at the infrastructure level will be vital to earning user trust and enabling fully autonomous agent ecosystems.

• Agent Frameworks & Tools – Agent frameworks (e.g., ElizaOS, G.A.M.E. by Virtuals, RIG, Heurist, REI) are development environments and libraries for building AI agents without starting from scratch. They provide the core architectures for agent “brains” – handling memory, decision-making, prompt responses, and task execution. While frameworks provide the brains of agents, there are onchain agent toolkits (e.g., Coinbase AgentKit, SendAI) that pre-package these frameworks for specific use cases and connect agents with smart contracts, wallets, payment rails, and onchain data. By using these frameworks and tools, agent developers can spin up robust agents with built-in support for advanced, multi-platform interactions, long-term memory, and onchain connectivity.

• Agent Launchpads – Platforms in this category help create, launch, manage, and/or monetize AI agents by packaging them as onchain entities (often with their own tokens). In practice, agent launchpads (e.g., Virtuals, auto.fun, ARC) let creators deploy new agent instances and bootstrap communities or funding around them. By aligning incentives (through tokens or fees), these launch platforms empower agent developers to sustain and scale their onchain agents as standalone projects or businesses.

• Multi-agent Coordination – Not every problem is best solved by a single agent. Multi-agent coordination protocols (e.g., Virtuals ACP, Questflow, Theoriq) orchestrate multiple AI agents (an “agent swarm”) working together to accomplish complex tasks. One agent might handle data gathering while another evaluates results, all overseen by an onchain orchestrator agent. This swarm approach can outperform monolithic agents by leveraging specialization and parallelism. By enabling agents to cooperate, multi-agent coordination platforms can expand the scope of what onchain AI can automate, from multi-step workflows to entire autonomous organizations.

• Model Context Protocols (MCP)  – Sitting at the intersection between AI agents and external data, model context protocols (originally created by Anthropic) have emerged as a key service. These protocols help standardize how onchain agents fetch relevant context, knowledge, or tools from outside sources. Instead of custom integrations for every data source or smart contract, an agent integrated with an MCP standard can plug into any compliant context provider (whether that’s onchain data, offchain databases, or web services) and retrieve the information or tools it needs. Decentralized MCPs such as Heurist and DeMCP provide agents with self-developed and open-source MCP services, enabling one-stop access to mainstream large language models, thus making onchain agents more adaptable and powerful in practice.

• AI App Stores - AI app stores (e.g., Alchemist AI, ARC Ryzome) consist of platforms that function as marketplaces and discovery layers for onchain agents, tools, and experiences. These app stores make it easy for developers to launch, monetize, and distribute agents or AI modules, while allowing users to browse, summon, or customize agents with familiar interfaces. These app stores serve not just as distribution hubs, but as coordination surfaces for the broader onchain AI economy, fostering interoperability between agents, tools, and protocols. As the number of onchain agents and AI-native applications grows, these platforms may become vital ecosystems in their own right—curating experiences, routing users, and capturing a share of the value flowing through agentic interactions.

Types of Agents

With rapid advancements in the agent infrastructure and services layers, we believe we can broadly categorize onchain AI agents into a few segments today:

• Trading / DeFi Agents – These agents specialize in financial actions – for instance, executing trades (e.g., Bankr*, Cliza), providing liquidity (e.g., BasisOS), optimizing yield (e.g., ARMA*, Mamo*), or arbitraging price differences in DeFi. They may also participate in prediction markets (e.g., Billy Bets*) or manage entire investment funds or portfolios (e.g., ai16z, aiXCB). These trading agents can react faster than humans, operate 24/7, and potentially make more data-informed decisions, which could enhance market efficiency (or perhaps outcompete human traders).

• Service Agents – Service agents are those that provide useful services to users or protocols. For example, an agent could provide relevant market analysis research and insights (e.g., aiXBT, BitQuant*, Chaos AI*). Some agents may handle DAO governance tasks – reading proposals, summarizing them, even casting votes according to a preset logic. Other service agents might audit smart contracts for bugs or auto-generate code for new smart contracts based on natural-language inputs (e.g., AgenTao, Kolwaii). There are also commerce-related service agents (e.g., Byte AI), which we’ll cover later, like agents that negotiate deals or pay for items on behalf of human users. In essence, these are the “autonomous workers” of crypto, automating onchain tasks that normally require human labor or attention.

• Entertainment Agents  – These agents focus on engaging with users. In gaming, AI agents act as NPCs (non-player characters) that can interact naturally with players. Unlike scripted game bots, these AI NPCs can learn and evolve, making games more immersive. Beyond games, we have social agents: think AI influencers (e.g., Luna) on platforms like X or Farcaster* that can post content and interact, or AI agents that can create artwork and IP based on community inputs (e.g., Botto). In the future, you might follow an AI influencer onchain who manages its own treasury (perhaps it earns crypto from content it creates on Zora8 or from tasks it does for followers). There’s also AI companions that provide hyper-personalized interactions, some with very nuanced, multi-modal expressions and actions (e.g. Nectar AI).

It’s early days, but these categories show the breadth of possibilities. From AI fund managers to AI friend sims, onchain agents could occupy many niches. What unites them is that they use cryptographic primitives as their playground and toolkit – holding assets, executing smart contract code, and leveraging the transparency and composability of decentralized networks. 

Apps

Hand-in-hand with autonomous agents, we’re also witnessing a wave of AI-powered onchain applications. These are apps and platforms that embed AI into their user experiences or core functionality. Below are a few domains where Onchain AI apps are taking shape:

• DeFi (“DeFAI”) – AI is making inroads into DeFi in multiple ways. One clear trend is AI-assisted trading and portfolio management. Instead of manually navigating complex DeFi protocols, users can leverage AI interfaces that handle it for them. For example, HeyElsa* is an AI-powered crypto co-pilot where users can simply ask its agent to perform tasks (e.g., “Swap X for Y”), and the agent will execute those actions across protocols. Protocols like Giza* provide access to non-custodial agents that can monitor DeFi markets, identity yield optimization opportunities, and dynamically manage positions with real-time market awareness. We believe this kind of AI-powered UX represents a “Wealthfront for Crypto” moment, where the robo-advisor is an onchain AI agent that’s purpose built for DeFi, effectively becoming a personal crypto portfolio manager that’s available to anyone.

• Gaming & Agentic Metaverses – Gaming is a natural playground for AI agents, and when combined with true asset ownership onchain, we get the concept of agentic metaverses. These are game worlds or virtual environments populated by AI agents alongside other agents or human players to create richer, more dynamic gameplay content. The agents could be friendly NPCs, autonomous adversaries, or even AI avatars controlled by other players. For example, Youmio is building an autonomous world where AI agents continuously learn, play, and entertain in real-time, creating ever-running simulations onchain. Additionally, firms like Farcade* are creating an AI-powered onchain gaming studio where anyone can “vibe-code” and distribute onchain games via natural-language prompts.

• Consumer AI  – AI is transforming consumer applications by making them more personalized, interactive, and intelligent. ChatGPT alternatives such as Venice and FreedomGPT allow users to access powerful models with privacy-preserving and censorship resistant inputs and outputs. In onchain social networks, AI agents can act as influencers, curators, or creators—managing content in feeds, generating posts, engaging in conversations, and even executing onchain actions (e.g., Clanker). Within onchain consumer apps (e.g., Zo), AI can help streamline onboarding, recommend actions based on onchain behaviors, or negotiate on behalf of users in peer-to-peer marketplaces. Lastly, AI companions (e.g., Nectar) allow users to create and interact with agents that respond with nuanced, multimodal expressions and actions – all verifiable onchain. These kinds of agentic experiences can help unlock a significant step-function improvement in crypto UX, thus bringing it closer to mainstream consumer expectations.

Commerce

One of the most profound implications of onchain AI is how it enables a new form of digital commerce - what Coinbase Ventures refers to as Agentic Commerce. This is commerce driven by AI agents transacting with each other and with humans. In such an economy, crypto becomes the preferred payment rail for machines as well as people​. The reasoning is straightforward: autonomous AI agents operating around the globe can’t walk into a bank, but they can trustlessly send and receive cryptocurrency on public blockchains. Crypto’s borderless, programmable nature makes it ideal for machine-to-machine payments, microtransactions, and automated contracts​. For example, the Coinbase Developer Platform team recently launched x402, a new open-source payment protocol that enables AI agents and applications to autonomously pay for GPU compute, API access, digital content, and more using crypto rails. Additionally, startups like Payman* and Skyfire* are building infrastructure services that enable agent-to-human or agent-to-agent payment orchestration using stablecoins like USDC.

While agentic commerce is still nascent, we believe it holds the promise of automating and accelerating business transactions in ways not seen before. Commerce could become a machine-speed, always-on affair, with agents negotiating deals, executing contracts, and exchanging value in seconds. Importantly, humans set the goals and parameters, and the agents do the rest. Blockchain’s role is to provide a safe and interoperable playground for these agents to transact – with clear rules (smart contracts) and reliable money (stablecoins).

Future Outlook

Looking ahead, the future of onchain AI appears full of potential, though it will unfold in stages. In the near term, we expect continued experimentation with onchain AI agents and AI-powered apps. Long-term, we believe crypto is poised to become the de facto economic layer for AI, meaning any advanced AI agent will use crypto to store value and settle transactions. As AI increasingly aids in writing software and smart contract code, the pace of innovation within the onchain economy could rapidly accelerate, bringing an influx of new applications and users.

However, realizing this vision comes with several challenges to consider. Agentic technology is still early, and some expectations have run ahead of reality. AI agents today are limited in reliability and scope, and it may take time before they can fully handle open-ended tasks safely. Scalability of blockchains will also be tested if swarms of agents start transacting simultaneously. There’s also a pressing need for new trust and governance frameworks. AI agents can supercharge onchain systems, but also amplify security and trust issues if not properly governed. 

From a value capture perspective, we believe unlocking the potential for the onchain AI economy will require the following: robust infrastructure that improves agent intelligence (e.g., data networks and post-trained models purpose-built for onchain agent use cases), services and tools for orchestrating agent behavior (e.g., multi-agent coordination, decentralized MCP, agent identity/payment rails), and distribution of agents to mainstream consumers (e.g., agent launchpads, AI app stores, and consumer AI). 

​​In conclusion, the rise of onchain AI represents a frontier for machine-driven intelligence. From autonomous agents executing smart contracts to onchain apps that adapt to user needs in real-time, this movement could redefine how humans and machines interact. It is an exciting era – one that Coinbase Ventures and many in the crypto community believe could lead to the next big leap in the evolution of the internet, an Agentic Web that ushers in a more autonomous and intelligent digital economy. 

Thanks to Hoolie (Coinbase Ventures), Luca (Base), Lincoln (Coinbase), Vik (Coinbase), Daniel (Variant), Josh (Contango Digital), Anand (Canonical), Teng (Chain of Thought), and EtherMage (Virtuals) for their thoughtful feedback and discussion on this post.

Related work

• Demystifying the Crypto x AI Stack, Coinbase Ventures, October 2024

• Blockchain for AI, Rajarshi Gupta, March 2024

Disclosures

The contents in this post — and available on any associated distribution platforms and any public Coinbase Ventures online social media accounts, platforms, and sites (collectively, “content distribution outlets”) — should not be construed as or relied upon in any manner as investment, financial, legal, tax, or other advice. You should consult your own advisers as to legal, business, tax, and other related matters concerning any investment. Investments in startups and early-stage businesses are inherently risky and highly illiquid. These types of investments are suitable only for qualified, sophisticated investors who can afford to bear the risk of losing the entirety of their investment capital.

Under no circumstances should this post or any other information provided on the Coinbase Ventures website — or on associated content distribution outlets — be construed as an advertisement, solicitation, invitation, recommendation, offer or inducement of any kind. Nor should this post or such information be construed as an offer to provide investment advisory, investment management, financial advisory, or financial management services. The information on the content distribution outlets should not be relied upon as research, investment advice or a recommendation of any kind.

Past results of Coinbase Ventures’ investments or investment strategies are not indicative of future results.  Coinbase Ventures may in the future have a position or holding in any company referred to on the Coinbase Ventures website, any associated content distribution outlets, or otherwise.