Analyzing A16z's "AI × Crypto" investment thesis and listing 14 crypto projects needed for AI Agents

Author: Stacy Muur, Crypto Researcher

Translation: Felix, PANews

Based on a16z’s three investment principles of “AI × Crypto,” Crypto Researcher Stacy Muur points out that the future of AI is not solely about enhancing intelligence but about how it integrates into the human economy. In this process, blockchain is an indispensable infrastructure. Below are the details.

As AI Agents begin to think, act, and trade independently, the core question becomes: how to enable AI to participate in economic activities safely. Blockchain can provide the necessary coordination layer, making autonomous Agents credible economic entities.

This article analyzes a16z’s “AI × Crypto” investment thesis: “Know Your Agent” (KYA) and how cryptographic trust enables AI Agents to collaborate. Additionally, it discusses why micro-payments are vital for a sustainable AI economy and which projects and infrastructure are worth关注.

Point 1: Blockchain as an efficient infrastructure layer for AI models and Agents to collaborate.

AI is gradually evolving to solve problems that only a few experts could previously address. Recently, ChatGPT 5.2 successfully solved a math problem that only a few hundred people worldwide could solve.

In the past, AI was often criticized for frequent “errors.”

But with AI’s progress, these “mistakes” can help it brainstorm like humans, combining ideas and establishing connections. To unleash this creativity at scale, it’s necessary to go beyond single models and build layered systems. In such a system, one AI system freely generates ideas, a second critiques them, a third refines the best parts, and a fourth verifies the final results.

However, once multiple AI systems operate collaboratively, two fundamental issues arise:

Interoperability

Accountability

Different models have various formats, lacking shared language or control layers, making coordination very difficult. When one AI proposes an idea, another improves it, and a third verifies it, it’s hard to determine who deserves credit, who should be rewarded, or who is responsible for errors.

Cryptocurrency and blockchain can precisely address this issue. They are not just intelligent systems but provide infrastructure that records who did what, when it happened, and each contributor’s share. Through verifiable logs, hashes, proofs, and automated incentive mechanisms, cryptographic technology can serve as a bookkeeping and coordination layer, enabling different AI systems to collaborate.

Focus List

1. Covalent: Building a modular data architecture that allows AI Agents to collaborate using shared, verifiable blockchain data. Multiple Agents can use its AI Agent SDK and “Zero-Employee Enterprise” workflows to complete complex tasks, while Block Specimens and GoldRush API ensure interoperability between blockchain and tools. This makes blockchain the foundation for data availability, verification, and incentive mechanisms.

2. Allora: Developing a decentralized coordination layer enabling multiple models to collaborate on very specific tasks for better results. Allora leverages cryptography to coordinate participation, verify contributions, and ensure AI Agents work together in a way that makes the system smarter over time.

3. Questflow: Building an on-chain orchestration layer where autonomous AI Agents can communicate, coordinate actions, and complete entire workflows together, rather than each Agent executing isolated tasks. Questflow’s Multi-Agent Orchestration Protocol (MAOP) allows Agent clusters to work together for reasoning, decision-making, action, and payment settlement.

4. Gaia: Providing routing, load balancing, and request services for a large number of independently running AI Agents. Using standardized runtime environments (WasmEdge), OpenAI-compatible APIs, and Agent combinations (LLMs + RAG + tools), Gaia addresses large-scale interoperability among heterogeneous Agents. The network has over 700,000 nodes and more than 29 trillion inference throughput, demonstrating its potential in real-world applications. Gaia does not rely on provider trust but uses protocol-level mechanisms (on-chain IDs, hosted contracts, staking) to introduce accountability for AI agent execution.

5. Sentient: Building the GRID open intelligence network, where 100+ models, Agents, data sources, tools, and compute providers work together as a single system. GRID routes each query to the most relevant specialized Agent and merges outputs into coherent results.

The network is live, with over 110 partners, and adopts a token-based model that rewards valuable outputs through staking and actual usage, aligning funding with utility. By allowing Agents to trade directly in $SENT, cryptography becomes the coordination and incentive layer for sustainable, scalable open network intelligence.

In addition to these projects, there are two interesting research papers. If you want to learn more and explore these fields deeply, consider reading:

1. Emergent Knowledge Intelligent Systems (ISEK): ISEK proposes a collaborative structure where humans and AI Agents not only perform tasks but also discover, negotiate roles, assemble temporary teams, and settle results through a native protocol cycle (Publish → Discover → Recruit → Execute → Settle → Feedback). Trust, memory, and incentives are fundamental: Agents have verifiable identities (Agent cards / NFTs), multi-dimensional reputations, and value exchange via tokenized micro-payments based on work performance.

2. LOKA Protocol: Building a decentralized framework for trustworthy and ethical AI Agent ecosystems.

LOKA is an academic proposal aimed at governing large-scale AI Agent ecosystems. It introduces a layered architecture where Agents possess autonomous identities (DID + verifiable credentials), graph perception communication, and decentralized moral consensus mechanisms, enabling Agents to think about “what they should do,” not just “what they can do.” LOKA explores how to embed accountability and moral norms directly into protocols using on-chain logs, reputation-weighted consensus, and even post-quantum cryptography.

Point 2: AI Agents need identity, not just more intelligence. “KYA” is the missing factor.

Today, AI Agents are already functioning in the real economy. They make payments, book services, trade assets, negotiate deals, and operate critical financial infrastructure via APIs, robots, scripts, and automation systems. These Agents are smart enough to work normally; intelligence is no longer the barrier. The real issues are identity and trust. When an Agent makes a payment, places an order, or signs a contract, no one knows who owns these actions, what it can do, or who is responsible if something goes wrong. Therefore, websites and merchants typically block them with CAPTCHAs, IP bans, and bot protections.

The solution is “KYA.” Agents need encrypted identities and verifiable credentials, just as humans need legal identities. Each Agent must have a signing key to prove its creator, representative (individual, company, or DAO), its permissions, and responsibility in case of damages. These credentials explicitly define the Agent’s spending, transaction, and data access limits, clarifying accountability.

Focus List

1. Billions is building “KYA,” using the Agent JS SDK, enabling Agents to generate their own DID (Decentralized Identity), prove control via cryptographic signatures, and manage keys through a modular Key Management System (KMS), thus establishing Agent identity, accountability, and reputation. Over 2,372,153 users have already joined.

Through collaboration with Privado ID (formerly Polygon ID), Billions leverages zero-knowledge self-sovereign identities for cross-service, device, and protocol privacy verification. Its core is $BILL, a fixed-supply ERC-20 utility token fueling the trust economy, with a cycle of network growth → verification activity → revenue → on-chain buyback → supply reduction → value appreciation → network growth, combining actual usage with long-term value accumulation.

2. cheqd.io: Building trust infrastructure for the Agent economy, turning KYA into a tangible product. Through Agentic Trust Solutions, AI Agents gain verifiable DIDs, fine-grained credentials, permissions, and attestations, all anchored in tamper-proof trust registries.

Using MCP (Model Context Protocol) servers, Agents can read/write identities, issue and present verifiable credentials, and prove their creators, permissions, and trustworthiness.

3. Vouched.ID: Developing a security, accountability, and compliance-focused KYA tech stack. Via MCP-I (Model Context Protocol—Identity), Agents obtain verifiable cryptographic identities, human authorizations, context-based operational limits, and full audit trails.

This stack, combined with knowthat.ai (a public Agent reputation registry) and Vouched Agentic Bouncer (which intercepts unauthorized or impersonating Agents), ensures autonomous AI operates safely within regulated environments.

4. ERC-8004 (Trustless Agents): An Ethereum proposed standard (EIP), currently not finalized. Its main goal is to implement “KYA” at the protocol layer. It defines how AI Agents can have verifiable on-chain identities, reputation, and execution proofs, allowing users and services to determine Agent authorization and trustworthiness without relying on centralized platforms. This EIP is actively being designed and discussed by the Ethereum Foundation team, with contributions from Coinbase, MetaMask, and others.

Point 3: Blockchain enables real-time, usage-based micro-payments and nano-payments, automatically compensating creators when AI tools or content are used, ensuring fair and transparent revenue sharing.

AI tools like ChatGPT, Claude, and Copilot ease user interactions but quietly disrupt the open web’s revenue model. The web relies on ads, subscriptions, and paid browsing to sustain itself, but AI fundamentally changes the value cycle:

Before AI: User searches → clicks website → website profits.

Now: User asks AI → AI reads website → provides answer → website traffic and revenue decline.

This creates a “hidden tax”: AI consumes information without paying content creators. If this continues, websites lose traffic, ad revenues plummet, creators stop publishing, and the open web shrinks. Ironically, this deprives AI of fresh, high-quality data. Laws can intervene, but progress is slow; thus, technical solutions aligned with incentive mechanisms are urgently needed.

The solution is usage-based compensation, where each AI use of information automatically rewards the content creator in real-time. Content is paid per AI usage (like Spotify per stream or YouTube per view), rather than fixed licensing agreements.

This is achieved through micro-payments and nano-payments, with AI attributing answers to multiple sources and using mathematical algorithms to proportionally distribute small rewards, rather than manual allocation. For example: Site A contributes 20%, Site B 30%, Site C 50%, and payments are split accordingly.

Blockchain and cryptocurrencies play a key role here; through smart contracts, automated payments are embedded directly into the network, allowing AI to continue providing services while fairly compensating the creators it depends on.

Focus List

1. Catena Labs: Building AI-native financial institutions designed for direct AI participation in the economy. Using open-source Agent Commerce Kit (ACK), providing wallets, verifiable identities, payment channels, and rule-based spending controls, enabling Agents to autonomously handle payments. ACK supports stablecoin payments, micro-payments, and inter-Agent transactions on blockchain testnets, allowing Agents to automatically compensate other Agents or human creators when using data, content, or services.

2. x402: Embedding micro-payments directly into standard HTTP requests with near-zero friction, enabling AI Agents to instantly pay for content, APIs, and compute costs. KITE AI upgrades this payment primitive into a full execution layer, creating a blockchain that allows autonomous AI Agents to reliably settle on-demand paid transactions at scale. KITE enables AI Agents to use x402 flow, native identities, and stablecoins for settlement, automatically paying creators, services, and other Agents at the point of consumption.

3. Alsa: Building a native AI payment and billing layer, where AI Agents pay only when executing actions, using a single account, tokens, and APIs. It supports tokenized on-demand micro-payments, backed by low-latency blockchain infrastructure and emerging agent-side payment standards.

Currently processing over 10.5 million x402 transactions (about 16% of network activity, mainly on Base), with plans to expand to Solana and Polygon, demonstrating that native AI micro-payments can operate reliably at scale.

Related: a16z: Five reasons blockchain is a key piece of the AI puzzle—from identity to payments