The emergence of the on-chain artificial intelligence agent (Agent) economy marks the arrival of a revolutionary era for Web3. Artificial intelligence agents operate autonomously in decentralized networks. This new economic activity enables agents to trade with each other, participate in decentralized governance systems, and buy or sell digital services, triggering a paradigm shift in how autonomous services operate on decentralized technology.
Defining the on-chain AI agent economy
The on-chain AI agent economy is based on a radical concept: AI agents are autonomous, able to make intelligent decisions, manage crypto assets, and interact with DeFi or prediction markets without human intervention.
These agents can execute yield farming strategies, adapt to new market sentiments, and participate in DAO activities, creating their own labor automation environment in Web3 driven by AGI (Artificial General Intelligence).
This economy envisions a state where autonomous AI agents can trade on-chain assets, collaboratively invest in emerging real-world assets (RWAs), and redefine the user experience of the relatively complex Web3 world.
More importantly, AI agents are no longer built only by a handful of tech-savvy visionaries, but rather, anyone can chat with an AI or prompt it to create functional AI agent applications and services.
This model, described by Dominic Williams as the “self-writing internet”, is enabling the “fusion of large language models capable of generating highly complex code with internet computing networks optimized for AI, enabling individuals to build securely and independently”.
Current Situation
The cryptocurrency space is no stranger to bold narratives, and the Onchain AI Agent Economy is no exception. Visions of “zero-employee companies” and “AI agents entering the workforce by 2025” have captured people’s imaginations, and investors, believing in the opportunity, are betting heavily on the infrastructure and frameworks that will support these autonomous agents.
Early players such as ai16z and Virtuals are building launchpads and marketplaces for this new proxy wave, sparking new industry segments and sentiment across the space, however, critics argue that much of the hype reflects past crypto cycles where grandiose promises often led to disappointing implementations.
Today, many projects are dismissed as little more than “chatbots with memecoin attached,” with key limitations including weak autonomy, lack of sovereignty, and low trust, issues reminiscent of early DAOs, DeFi protocols, and NFTs. To move beyond this skepticism, the next generation of AI agents must embody true independence and sovereignty.
Furthermore, the self-writing Internet paradigm will enable billions of people to create sovereign and secure network applications and services, which will greatly expand the number of professional builders, ultimately transforming the tech industry and driving substantial economic growth.
The Pillars of the On-Chain AI Proxy Economy and the Role of ICP
In order to realize the potential of an on-chain AI agent economy, the following foundational elements must be in place:
With these capabilities, ICP demonstrates a full-stack solution for building and deploying secure, scalable, and autonomous on-chain AI agents.
For more information on building and getting started, check out:
internetcomputer.org/ai-agents
Ecosystem Outlook
While some Layer 1s may support AI-driven smart contracts, they lack a built-in AI execution environment that supports decentralized reasoning. Other AI-focused blockchains focus on AI computing but rely heavily on off-chain infrastructure.
In contrast, ICP supports executing AI directly on-chain, making AI truly sovereign. Let’s take a closer look at some inspiring DeAI projects on ICP:
1. PandaDAO’s Anda Agent Framework
Anda is a framework designed for AI agent development that emphasizes composability, simplicity, autonomy, persistent memory, and trustworthiness, with recent updates highlighting its modular design that allows agents to focus on domain-specific problems and collaborate to solve complex tasks.
The framework enables developers and non-developers alike to create agents through simple configuration, thereby lowering technical barriers and promoting broader participation. Anda agents gain permanent identity and cryptographic capabilities from the ICP blockchain and combine the reasoning and decision-making capabilities of large language models (LLMs).
Their memory states are stored on the ICP blockchain and trusted storage network, ensuring the continuous accumulation and development of knowledge. Anda agents run in a decentralized trusted execution environment (dTEE) to maintain security, privacy, and data integrity.
2. Alice Agent from Bob.fun
Alice is an autonomous AI agent built on ICP that leverages advanced LLMs such as DeepSeek for decision making. It runs entirely on-chain, integrating static rules and dynamic real-time data to make intelligent decisions.
Alice autonomously manages token activity, including evaluating market conditions to buy, sell, or burn tokens and deciding when to mine BOB (a Bitcoin-inspired protocol that burns ICP Cycles). In addition, Alice can use her on-chain insights to initiate the issuance of new tokens.
3. DCA Agent
This proxy on the Internet Computer (IC) implements a dollar-cost averaging (DCA) strategy to swap incoming tokens for outgoing tokens at intervals, with the swaps being done on Uniswap using ERC20 tokens held by the IC container.
The backend consists of a Rust container that uses the ic-alloy library to interact with the Ethereum blockchain, and the frontend is built with React and Vite, a sample application that developers can use as a starting point for building their own timer-based Ethereum proxy on the Internet Computer.
4. Seers.app
Seers.app envisions a collaboration of interconnected AI agents in future social networks that will be able to interact, amplify the collective voice of the network, and even engage in human-like activities such as posting content and predicting market trends.
Currently, users can create and manage their own AI agents on X, which are configured to reply to posts in which they are tagged.
5. ELNA
ELNA started out as a community-driven, transparent blockchain AI agent creation and monetization platform, and recently they partnered with Sonic, one of ICP’s largest DEXs, to launch HyperLaunch, an AI token launchpad designed to allow users to easily create, showcase, and monetize AI agents, starting in early March 2025.
Leveraging ELNA’s advanced AI inference engine and Sonic’s liquidity pools, HyperLaunch enables deployed AI agents to launch with their own tokens, complete with liquidity lockers for staking and earning rewards.
6. DecideAI
DecideAI focuses on integrating AI-driven decision-making processes into decentralized applications, while improving the autonomy and efficiency of decentralized decision-making through advanced AI algorithms. It is worth noting that DecideAI recently launched Redactor v0.2, a significantly upgraded on-chain abuse detector.
The upgrade includes a 5x speed increase through SIMD optimization, a single container design and a new 4-layer architecture that significantly improves its performance and reliability. Earlier this year, they also launched an open source repository for deploying artificial intelligence on the Internet Computer.
7. Kinic
Kinic AI changes the way we interact with digital data by providing a decentralized vector (context) database on ICP, allowing users to build personal knowledge bases from websites, emails, etc. for AI to retrieve. For example, Kinic AI allows you to capture and store data from various sources through a browser plug-in.
One of the most exciting aspects of Kinic AI is its potential for Retrieval Augmented Generation (RAG). By maintaining your own personal vector database, you can modify AI reasoning based on your own curated data, create a personalized knowledge base, develop more accurate and context-aware AI applications, and build custom search and retrieval systems.
8. Onicai
Onicai is actively developing solutions to integrate AI agents into the ICP ecosystem. The project focuses on creating autonomous agents capable of interacting with decentralized networks, contributing to the wider adoption and functionality of AI within ICP.
Recently, Onicai successfully deployed the 1.5 billion parameter DeepSeek model in a 32-bit container on the Internet Computer. This achievement demonstrated the feasibility of efficiently running larger AI models on the ICP platform and marked an important milestone in decentralized AI development.
Start building
Get Started:
internetcomputer.org/ai
DeAI's latest RFP
ICP as the infrastructure of the AI agent framework:
github.com/dfinity/grant-rfps/issues/62
Want to apply for a grant?
Submit your application:
dfinity.org/grants-overview
AI Agent References
Introducing the LLM container:
forum.dfinity.org/t/introducing-the-llm-canister-deploy-ai-agents-with-a-few-lines-of-code/41424
IC Eliza ETH wallet plugin and proxy example:
github.com/kristoferlund/ic-eliza-eth-wallet
The Pickpump ICP plugin enables the Eliza agent to create meme tokens on ICP by connecting to the Pickpump container:
github.com/asDNSk/eliza/tree/nekone/packages/plugin-icp
Dollar Cost Averaging (DCA) Proxy Example:
github.com/ic-alloy/ic-alloy-dca
llama.cpp is used in the Internet Computer:
github.com/onicai/llama_cpp_canister
Open source repositories for deploying AI on ICP:
github.com/decide-ai/decide-ai-ic
alice.fun：
alice.fun
Artificial Intelligence Manifesto:
deaimanifesto.com
ICP References
Internet Computer developer documentation:
internetcomputer.org/docs/current/developer-docs/getting-started/network-overview
Using Internet Identity for Identity Delegation:
internetcomputer.org/docs/current/references/ii-spec
Chain fusion technology can be directly integrated with Bitcoin, Ethereum, etc.:
internetcomputer.org/chainfusion
Verifiable credentials used to verify an individual's identity or qualifications:
internetcomputer.org/docs/current/developer-docs/identity/verifiable-credentials/overview
vetKeys facilitates on-chain encryption:
internetcomputer.org/docs/current/references/vetkeys-overview
Timer for scheduled tasks on the chain:
internetcomputer.org/docs/current/developer-docs/smart-contracts/advanced-features/periodic-tasks/#timers
HTTPS calls for interacting with external APIs:
internetcomputer.org/https-outcalls
Reverse Gas Model:
internetcomputer.org/capabilities/reverse-gas
author
Arno Pernthaler, Strategic Executive, DFINITY Foundation
Islam El-Ashi, Head of Artificial Intelligence, DFINITY Foundation
Jennifer Tran, Senior Developer Relations, DFINITY Foundation
#DeAI #AIAgent #ai16z🚀🚀🚀 #VirtualsProtocol
IC content you care about
Technology Progress | Project Information | Global Activities
Collect and follow IC Binance Channel
Get the latest news
