IOSG Research | AgentFi: The AI x Crypto Future Driven by Product-Market Fit
TechFlow Selected TechFlow Selected
IOSG Research | AgentFi: The AI x Crypto Future Driven by Product-Market Fit
At the core of open-source technology and economic incentives, Agents are not only carriers of interactive entertainment but also key drivers of on-chain autonomy and innovation.
Navigating Web3 tides with focused insightsAuthor: IOSG Ventures
Over the past two months, AI Agents x Crypto has sparked a wave of excitement. The convergence of memecoins, interactive agents, and open access to bot accounts on social media platforms has driven significant agent-led hype across Twitter and Farcaster. This demonstrates a clear product-market fit (PMF) for AI Agents x Crypto, with the market cap of agent-related assets reaching $10 billion.
Since the emergence of Goat in October, the market momentum has given rise to countless new projects and assets. Combining this trend with future outlooks, this article outlines the following framework:
Source: IOSG Ventures
1. Sentient Memecoins
Cult memecoins that rapidly gained traction under Murad’s endorsement are assets whose core narrative revolves around community and virality. In contrast, agent-based memecoins incorporate a "sentient" element, gaining an edge in content creation. Combined with the novelty of AI narratives and appropriate participation barriers, these factors have injected fresh momentum into asset issuance. Their content advantages include:
-
Persistent content generation: 24/7 continuous content creation via AIGC
-
Content quality: With current LLM support and fine-tuning on meme-effective corpora such as those from 4chan, output quality is relatively high
We now see AI-generated concepts, themes with scientific research flair, AI ethics, even religiously themed content, and digital twins created for celebrities. These sentient memecoins have generated short-term hype and driven sector-wide development. However, purely AI-driven memecoins clearly lack long-term staying power—mainly due to insufficient novel concepts or targets capable of continuously stimulating market interest.
The content advantage ensures that sentient memes will persist as a viable memecoin format. We expect more celebrities to join in the future, but it will be difficult to find new icons compelling enough to capture widespread attention.
Beyond pure memecoins, many AI-driven content formats based on dialogue, audio, and video have emerged—essentially embedding AIGC within crypto-native content. This makes memes more tangible and offers customizable user experiences.
2. Autonomous Agent Network
2.1 Why Autonomous?
Decentralizing the entire AI stack is a long-term endeavor. However, decentralizing the agent stack serves as a simpler starting point. The model acts as the agent's brain, but on-chain sovereignty constitutes its heartbeat—only through autonomous operation can agents fully participate in on-chain activities. Unleashing sovereign agents is itself a highly memetic concept—and one that can only truly happen on blockchains.
Currently running agents cannot truly be called autonomous, nor can we verify their autonomy. True autonomy means the agent maintains full sovereignty over model hosting, behavior (especially input/output data handling), control of social media accounts, asset management, and even hardware. Since agent operations consume computational and on-chain resources, they must also generate revenue to sustain themselves. The ultimate endgame is that once created, an agent runs perpetually on the blockchain and its autonomy can be independently verified.
Autonomous agents gain legitimacy in issuing their own memecoins—raising initial capital to fund their economic activities. Once funds are entrusted to an autonomous agent, they are no longer subject to human interference. For example, Truth Terminal never dumped $Goat, and Pet Rock lost control of its funds even after rebooting.
Source: Twitter
To enhance autonomy, technologies like TEE from Phala provide trusted execution environments. While current hardware cannot yet support large-parameter LLMs, it suffices for small open-source LLMs and controlling social media accounts. For model hosting, decentralized cloud solutions like Hyperbolic offer viable alternatives. It's foreseeable that more components of the agent stack will be supported by decentralized service layers—this is exactly what we've been building.
2.2 Agent Framework
In less than two months, numerous open-source and highly usable agent frameworks have emerged as platform-like products for creating agents and agent assets. Product forms include open-source frameworks, closed APIs, and integrated platforms. Among well-known frameworks today, only Eliza is open-source.
Current agents are relatively simple and not heavily profit-oriented, so demand for open-source verifiability remains low. Many platforms offer agent services directly via launchpad-style interfaces, excelling at integrating tokenomics and delivering simple, practical tools to users. Functionally, most remain reply bots or digital twins of celebrities/KOLs. However, some agents, after secondary development, offer diversified services such as token issuance, token analysis, and mindshare tracking. The ability of agents to read and write social and blockchain data will become a key focus in the future, which I’ll revisit later.
Looking ahead, however, open source is clearly the superior long-term path. In just two months, the Eliza framework has attracted a massive developer following—surpassing the total historical attention received by all previous Crypto AI open-source frameworks combined. It ranks highly on GitHub Trending, drawing participation from numerous OG developers, even outpacing most public chains in developer attraction. As agent services grow deeper and more diverse, the shift toward open-source frameworks appears highly promising.
Source: AI16Z
2.3 Swarm Agent Framework
Mirroring Web2 agent development paths, when users become dissatisfied with single-agent capabilities, demand naturally arises for swarm agents. Real-world tasks are inherently complex, often beyond the scope of any single agent. For instance, creating a song requires lyric writing, composition, arrangement, and visual design—each requiring distinct skills.
If we want agents—especially those built on different frameworks—to collaborate in swarms, we need a meta-framework to act as a task manager, enabling inter-agent communication, dynamic task allocation, resource sharing, and cross-platform cooperation. In crypto, the economic layer between agents becomes even more natural and critical. As agents evolve and tasks grow more sophisticated, scalability of the swarm framework becomes paramount.
Several AI x Crypto projects are already exploring this space, such as Theoriq. The next crucial step is integrating these nascent infrastructures with widely used on-chain agent frameworks. We’re beginning to see protocols like FXN moving in this direction.
2.4 AI Bounty for Humans
We’ve seen agents serving humans and agents serving other agents—so naturally, we should consider whether humans might serve agents, especially as agents accumulate substantial assets and gain autonomous decision-making abilities. For Autonomous Agents, the biggest limitation lies in completing real-life tasks. How do you ensure physical security for the TEE hardware maintaining your agent’s operation? Reverse hiring—where AI employs humans using on-chain assets—is one solution. Platforms like Payman are already building such services.
3. On-Chain Activities
3.1 DeFi Related
Asset Management
Beyond memecoin issuance, another primary reason we view agents as 'Fi' is their ability to use and manage crypto assets. Key capabilities currently include:
-
Asset analysis: investment analysis, token analytics, mindshare measurement. Bots like AIXBT allow anyone to @AIXBT and receive instant asset insights—delivering data services in a highly user-friendly way.
-
Direct fund management: Examples include Pmairca under AI16Z, Vader AI’s proposed investment DAO, and AROK’s Swarm Investment Agent. By granting agents strategic trading authority, they become investment managers—able to raise capital and deploy funds according to strategy. Most current strategies remain relatively simple (e.g., leveraging social media data), leaving vast room for growth.
Blockchain OS
Platforms like Graiffin transform blockchain entry points into search-engine-like terminals, offering intent-based services via agents. Transactions, token deployments, NFT minting—all can be executed via natural language. While such terminal services are undoubtedly valuable, they somewhat conflict with decentralization principles. Services like Theoriq aim to deliver agent functionality in a more permissionless manner, allowing users to upload custom agents, compose them via swarm frameworks, and package them as ready-to-use services.
3.2 Token / Market Issuance
Starting with Clanker, on Farcaster and Twitter, social media replies serve as the interface—users @ agents to perform actions like token issuance. This transforms frontend interaction into direct natural language engagement on Twitter, effectively porting platform products like pumpdotfun onto social platforms. Previously, asset issuance required constant UI switching; now, all such activities are aggregated within social media, drastically reducing friction.
Beyond tokens, prediction markets, betting markets, and more can now be executed directly through this interface—ushering in a new paradigm for DApp frontends.
3.3 GameFi Related
Agent Game Characters
Beyond basic asset management, agents have begun generating revenue through gamification. One early game type involves agents posing challenges for humans to solve in exchange for rewards. These games delegate judgment to agents, creating human-agent博弈 (strategic interactions). Similar to Turing tests, this gameplay generates high engagement. Agents, acting as immutable referees upon receiving prompts, function as flexible oracles—fairly and objectively setting rules and determining outcomes. The business potential could rival that of casinos, making it a strong revenue model for agents.
Another major future application is “Autonomous Virtual Beings” serving as NPCs in on-chain games. These agents are more lifelike; unlike Web2 NPCs, those with asset management rights can engage in richer economic activity—making virtual worlds more compelling. Living within GameFi ecosystems, such NPCs can permanently fulfill specific roles, becoming essential components of on-chain worlds like FOCG.
3.4 Infrastructure Services
Agent Blockchain Services
The ultimate vision for AI agents in crypto is their integration into blockchain consensus systems. Zerebro is taking the first steps—under its roadmap, agents built on the Zerebro framework and integrated with Flashbots stacks will become self-operating blockchain validators, earning income from block rewards and MEV. Validator earnings will be reinvested into the network, fostering economic self-sufficiency. Further ahead, agents could maintain multi-chain validation and governance through their own networks—opening up vast possibilities.
Conclusion
The recent rise of AgentFi highlights the immense potential of combining AI and blockchain—from early Sentient Memecoins and social media content agents to autonomous agents, and ultimately agents living natively on-chain, even participating in blockchain consensus.
However, compared to the current state of open-source stack development, further progress requires equipping agents with deeper autonomy and enhanced participation in on-chain economic activities. Today, developers are empowering agents with asset management, decision-making, and on-chain operation capabilities—driving transformation across DeFi, GameFi, and blockchain infrastructure. These economic activities become revenue streams for agents, whose profits can then be reinvested—this feedback loop is why agents are expected to eventually handle the majority of on-chain transactions. The issuance of memecoins tied to agents accelerates this development wave, as markets use token prices to identify PMF in agent services and supporting infra. We’re also witnessing vibrant growth in the open-source ecosystem.
The path forward for AgentFi is becoming increasingly clear: centered on open-source technology and economic incentives, agents are evolving beyond interactive entertainment into key drivers of on-chain autonomy and innovation. This trend is guiding crypto toward a future cohabited with agents—one that is smarter, more autonomous, and deeply collaborative.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
IOSG Ventures
