A simple explanation of how @ai16zdao's eliza framework works 🧵:
Web3 AI Agents Guide
Learn how to build, deploy, and manage web3 AI agents.
Table of Contents
1. Why Integrate AI with Blockchain?
In Chris Dixon's words,
“AI is going to end the internet as we've known it. Its advancements are inevitable and will be mostly beneficial. But AI is already upending the internet's economic covenant, and big tech companies are best-placed to reap the rewards. AI-powered deepfakes and bots are also degrading people's trust in the online world.”
To address these challenges without slowing down AI innovation, blockchains can:
- Enforce ownership for users and creators
- Verify identity to isolate bots and imposters
- Ensure authenticity to prevent tampering
Crypto can carry the torch of the open internet and help keep it creative, open, and diverse.“
2. What are AI agents?
AI agents are essentially autonomous software programs powered by LLMs that can understand, plan, and execute tasks with little human intervention. Think of them as digital assistants on steroids - they can handle more complex workflows, make decisions, and even interact with other agents to accomplish certain goals depending on the functions.
3. Use Cases of AI Agents in Crypto
AI agents can monitor smart contracts for potential issues or inefficiencies, suggesting optimizations or automatically executing adjustments based on new data or market conditions.
AI-driven trading bots can execute trades based on complex algorithms that learn from market conditions, offering strategies that adapt in real-time to maximize returns or minimize risks.
AI can assist in verifying identities on the blockchain, ensuring compliance with Know Your Customer (KYC) regulations while maintaining privacy through zero-knowledge proofs or similar technologies.
AI can enhance DeFi platforms by providing better risk assessment models for lending, yield farming, or token swaps, making these services more secure and user-friendly.
AI agents could manage cryptocurrency portfolios, rebalancing them according to market conditions, risk tolerance, or to achieve specific investment goals like diversification or capital growth.
AI agents can monitor listings, negotiate prices, and manage entire NFT collections. This automation can help in optimizing sales, managing collections, and even in the discovery of undervalued NFTs through data analysis.
In the governance of Decentralized Autonomous Organizations (DAOs), AI agents can analyze proposals, simulate potential outcomes, delegate voting powers, and facilitate community decisions.
4. Core Challenges of Web3 AI Agents
- Most agents are powered by centralized, closed-source models (like GPT-4, Claude)
- Even "open-source" solutions rely on centralized infrastructure (Groq, Together AI)
- Lack of true decentralization despite marketing claims
- Zero verifiability of agent actions
- Private key management risks
- Potential for unauthorized trades or fund mismanagement
- High vulnerability to prompt injections and LLM jailbreaks
- Data poisoning risks
- Most actions happen off-chain
- Limited differentiation in capabilities between agents
- Identical API calls and functionality across different agents
- Difficulty in handling complex, long-term objectives
5. Popular Web3 AI Agent Frameworks
Each framework has it's own unique features and capabilities. which one to use completely depends on your use case and the specific requirements of your project.
Framework | Memory | Focus | Capabilities | Chains | Edge |
---|---|---|---|---|---|
ElizaOS A TypeScript-based framework by AI16z with ~60% market share. Features extensive multi-agent simulation capabilities, cross-platform social engagement, and a thriving community of 6,000+ GitHub stars. Leading framework for creating AI agents that handle trading, social interactions, and community management. | RAG, multi-system memory | Social media, character, and trading agents | Text, voice, media interactions, image gen, Coinbase webhooks, search, NFT minting, TEE, wallet gen | Solana, Aptos, Flow, ICP, NEAR, Base,TON,SUI,Others | First-mover advantage, extensive GitHub community, robust plugin system, and broad model support |
RAG, Vector stores (Pinecone) | Social media agents with trustless transactions | Text, image + video gen, NFT minting, contract deployment, music gen, IP licensing | Solana, Base, Polygon | Strong Python ecosystem integration, specialized for creative and social media applications | |
RAG, Vector store DBs (MongoDB, LanceDB) | LLM-powered apps | Swarm intelligence, data pipelines, report generators, knowledge crawlers | Solana, Base | High performance, reliability, and modular design optimized for enterprise use cases | |
ERCData for permanent storage, Multi-layered memory systems | AI-blockchain integration, Pattern learning | Pattern recognition, Complex relationship mapping, Context preservation, Hierarchical data organization, Oracle Bridge integration | Ethereum, Base | Advanced data paradigm with ERCData, Oracle Bridge for real-world data integration | |
N/A | Cross-platform chatbots | Text, image gen, voice processing, wallet integration, payment system | Heurist Chain | Full-stack DeAI ecosystem, overhead cost reduction | |
GAME The Generative Autonomous Multimodal Entities framework by Virtuals (~20% market share) designed for gaming agents. Features 200+ projects, 150K daily requests, and rapid weekly growth. Includes built-in ERC6551 wallets, knowledge graphs, memory embeddings, and reinforcement learning. | RAG, LT and Working Memory | Gaming and social media agents | Text, voice, gaming, wallet, live-streaming, music gen | Solana, Sui, TON, zkSync, Others | API-driven architecture, no-code integration, strong gaming/metaverse ecosystem adoption |
On-chain storage | On-chain AI agents | On-chain operations, agent interactions | Sui | Fully on-chain agent operations | |
Dynamic data integration | Data-driven agents | Data gathering, model fine-tuning | Multiple chains supported | Continuous learning and evolution |
6. Build your first Web3 AI Agent
Eliza Framework
ZerePy Framework
ZerePy is now live github.com/blorm-network/… thanks to @ayoubedeth for a walkthrough video For our v1 we wanted to make it seamlessly easy to launch a personalized agent that can post on social platforms. The upcoming updates will be focused to expand agent capabilities,…
- ZerePy with EternalAI API Integration
Virtuals G.A.M.E. Framework
Another week, another tutorial. @celesteanglm just dropped some tips on how to configure your agent using the GAME sandbox! Check it out now:
Missed out on our session yesterday? We've got a video out where @celesteanglm walks you through all the functionalities of GAME, how to add custom functions and simulate outputs! Start building your first agent or improving your existing agent now:
Missed out on our session yesterday? We've got a video out where @celesteanglm walks you through all the functionalities of GAME, how to add custom functions and simulate outputs! Start building your first agent or improving your existing agent now:
The @goat_sdk from @crossmint now supports Game agents by @virtuals_io! Watch this @GAME_Virtuals agent swap USDC → $MODE on @modenetwork 👀
RIG Framework
Build with Rig, join the $arc complex. Let’s create an intelligent agent for crypto portfolio analysis across networks. Using: - Rig for orchestration -@zapper_fi for on-chain data - GPT-4 for generation Full code: github.com/Tachikoma000/r… (0/13)
7. Libraries to Connect Your Agent to Blockchains
Library | Description | Links |
---|---|---|
agentipy | A Python toolkit for connecting AI agents to Solana blockchain, featuring token operations, trading, LangChain integration, and DeFi capabilities | |
cdp-agentkit | Coinbase Developer Platform (CDP) Agentkit for Python - A framework-agnostic toolkit for bringing AI Agents onchain with features like wallet management, token operations, NFT deployment, and LangChain integration | |
goat | Framework for connecting AI agents to blockchains, supporting multiple agent frameworks (Langchain, Vercel AI SDK, Eliza), wallet types, and 30+ blockchains. Provides ready-made tools for token operations, DeFi interactions, and smart contract integration. | |
Hyperbolic-AgentKit | A template for running AI agents with blockchain and compute capabilities. Includes features for GPU operations and blockchain operations. | |
solana-agent-kit | Toolkit for connecting AI agents to Solana. Enables agents to perform 15+ actions including trading, launching tokens, lending assets, sending compressed airdrops, executing blinks, and launching tokens. |
Guides to use this libraries
How to create your Based Agent with 0 lines of code Recorded a quick demo video to show how easy it is and walk you through the entire process step-by-step All the links are in the comment below
Learn how to create your own AI Agent using Solana Agent Kit by @sendaifun in less than 60 seconds using Replit. It can perform on-chain actions using Solana Agent Kit. Checkout the video, replit link below
WANT TO MAKE A VIRAL AGENT?! my @vercel AI agent just sent me a penguin NFT for christmas without even knowing my wallet address?! build your own now with GOAT, @crossmint, @vercel, and @solana. under 10 minutes, i promise. links in next tweet 👇
We built the first AI agent that has its own computer powered by @hyperbolic_labs! AI agents are now GPU rich! We developed an AgentKit that allow AI agents to • Check GPU availability • Rent & manage GPU compute • Access & run commands on remote machines Why does this…
8. Infrastructure for Your Agent
Swarm Intelligence & Coordination
With more specialized agents entering the ecosystem, seamless communication between them becomes essential. Swarm intelligence allows agents to work together as a team, pooling their capabilities to achieve shared goals. Coordination layers abstract away the complexity, making it easier for agents to collaborate.
A multi-language AI Agents infrastructure leveraging Python, Solidity, RIDE, and TypeScript, designed for seamless integration of blockchain and AI capabilities. The project empowers the creation of AI agents that autonomously handle tasks like data analysis, content generation, and decentralized decision-making. With cross-chain compatibility (Waves, Base, BNB), it ensures interoperability and scalability for blockchain-native and off-chain AI applications.
Model Creators & Marketplaces
2025 will see an explosion of new AI agents, and many of them will be powered by decentralized models. These models will be more advanced, incorporating human-like reasoning, memory, and even cost-awareness.
Nous Research is a leader in the development of human-centric language models and simulators. They focus on model architecture, data synthesis, fine-tuning, and reasoning, aiming to align AI systems with real-world user experiences. Notably, they are working on a 'hunger' mechanism to introduce economic constraints to AI models, teaching agents to prioritize tasks effectively.
Data Providers
The quality, reliability, and integrity of data directly impact the performance of AI models. However, sourcing and labeling high-quality data is expensive, and bad data leads to poor results.
A decentralized indexing and querying solution for blockchain data, offering up to 100x faster indexing and 90% cost reduction compared to traditional RPC providers. Features a modular architecture with multiple products including SQD Network (distributed query engine), Squid SDK (TypeScript toolkit), SQD Cloud (PaaS), and SQD Firehose (subgraph adapter).
A decentralized data warehouse with sub-second ZK coprocessor for SQL, enabling trustless data processing for smart contracts. Combines comprehensive blockchain indexing with the ability to join onchain and offchain data through their Proof of SQL technology, allowing smart contracts to query complex data with ZK-proven results.
SQD.AI is an AI agent data infrastructure platform designed to support the demands of billions of autonomous AI agents. It provides a decentralized data lake that enables these agents to access and manage data independently. SQD leverages blockchain technology to ensure transparency and security, with its token ($SQD) serving as a means of accessing and rewarding participants in the ecosystem. The platform is designed to scale with the growing need for real-time, high-throughput data in the AI industry.
Verification & Security
Trust is the foundation of decentralized AI. As AI agents become more autonomous, we need systems that allow us to verify what's happening under the hood.
Compute Providers
The computational needs for AI are skyrocketing—doubling roughly every 100 days. Traditional cloud services like AWS aren't scalable for this demand, both in cost and accessibility.
Agents may utilize this API for private, uncensored inference (text, image, code).
iGam3 is a decentralized edge network designed to power AI agents by leveraging Zero-Knowledge (ZK) technology to ensure privacy and security. Built on blockchain infrastructure, it integrates AI processing with Decentralized Physical Infrastructure Networks (DePIN) to enable scalable, decentralized AI solutions. One of the core layers to this network, iGam3, aims to democratize access to AI and empower individuals and small groups to drive innovation.
AI-Agent Demo Days
Solana AI Hackathon Demo Day x.com/i/broadcasts/1…
all the AI Hackathlon Solana Projects without tokens launched in 1 place will leave link in the comments for anyone who may find this useful