Hook
Actually, the most important blockchain news this week isn’t on-chain. It’s Alipay’s AI open platform launch. They upgraded 10 million mini programs to be callable by AI agents via an MCP-like protocol. 10 billion users. One centralized gateway. And a commercial settlement layer that bypasses any decentralized token model.
Chaos is just data waiting for the right query. Here’s the query: What happens when the world’s largest payment platform turns its entire service ecosystem into an AI agent’s tool library?
Context
On December XX, 2024, Alipay announced the AI Open Platform. Merchants can upgrade their mini programs, APIs, and services so they can be invoked by large language models through a standard protocol—the article calls it “MCP” (likely Model Context Protocol or a proprietary equivalent). The platform’s AI assistant, named “Abao,” acts as the entry point, understanding user intent and routing requests to the appropriate merchant service.
Key features: cross-platform distribution (cars, AR glasses, smartphones), commercial settlement built-in, and a developer ecosystem that already powers millions of mini programs. This is not a prototype. It’s live, with open beta invitations.
From a blockchain perspective, this is the antithesis of decentralized agent coordination. But it’s also the most practical demonstration yet of what AI agents need: standardized service interfaces, payment rails, and user trust.
Core
Let me break this down with the forensic mindset I developed auditing ICO ledgers in 2017. Back then, I traced 14 wallet clusters hiding governance control. Today, I’m tracing a different kind of centralization—the service invocation layer.
The Protocol Trap
Alipay’s MCP is a centralized standard. It defines how any LLM (likely their own Tongyi Qianwen) can call any mini program service. Think of it as a proprietary API gateway with AI routing. The merchant doesn’t need to understand smart contracts or oracles. They just follow the MCP spec, and their service becomes an “AI Skill.”
On-chain, we have similar efforts: The Graph’s subgraphs, Chainlink Functions, and various “AI Oracle” protocols. But none have the scale Alipay is deploying. The difference is trust: Alipay’s trust is institutional, backed by regulatory compliance and a 10-year payment history. Blockchain’s trust is cryptographic, backed by code and consensus.
The Data Flow
Alipay’s AI platform will generate an immense dataset: user intents, service invocations, successful and failed transactions. This is the data flywheel that no current on-chain platform can match. Why? Because most on-chain AI agent platforms lack real economic transactions. They’re building infrastructure without the liquidity of actual commerce.
I’ve seen this before. In DeFi Summer 2020, I ran SQL queries on Dune showing 70% of yield was from arbitrage bots. The real value wasn’t the yield itself—it was the transaction volume that validated the protocols. Alipay’s AI platform will validate something similar: that AI agents can drive real commerce, not just token trades.
The Agent-Centric Standard
MCP is designed for agent-to-service communication. It defines intents, context windows, and response formats. This is eerily similar to the “Agent-to-Agent” (A2A) protocols emerging in crypto (e.g., Autonolas’s agent framework, Fetch.ai’s inter-agent communication). But Alipay’s version is closed, optimized for its own ecosystem.
The blockchain industry should study this. If centralized platforms define the standard for AI agents’ first billion interactions, decentralized standards will forever be playing catch-up. “Trust the hash, not the headline”—but the headline here is that centralized AI agents are eating the world while we debate governance tokens.
Liquidity Fragmentation or Concentration?
Some argue that Alipay’s move fragments the AI agent ecosystem across different platforms. I disagree. It’s concentrating liquidity of user attention and merchant services into one platform. This is the opposite of fragmentation. It’s a monopoly in the making.
On-chain, we complain about liquidity fragmentation across L2s. But we have the tools to unify—cross-chain messaging, shared sequencers. Alipay doesn’t need that. It has one database, one settlement layer, one trust model. The blockchain community should take notes: standardization of service interfaces matters more than decentralization of the sequencer.
Contrarian
Here’s the counter-intuitive angle: Alipay’s centralized MCP might actually accelerate decentralized AI adoption.
Several billion non-crypto users will now experience AI agents as a normal interface for everyday tasks—ordering coffee, checking insurance, paying bills. They’ll learn to expect agents that understand intent and execute transactions seamlessly. This expectation will eventually exceed what Alipay can offer, especially when users want to transact across different platforms (e.g., use WeChat Pay within Alipay’s AI agent).
Cross-platform agent communication demands open standards. Once users demand interoperability, centralized walled gardens will crack. The blockchain industry’s job is to have the open, permissionless protocol ready—one that allows any LLM to call any service, with settlement via stablecoins or L2s.
Based on my analysis of the Terra/Luna collapse in 2022, I learned that central points of failure eventually break under stress. Alipay’s AI platform will face regulatory, security, and scaling pressures. Privacy concerns alone could drive users toward decentralized alternatives where data remains local.
But—and this is the key—the blockchain industry must stop building for speculation and start building for real utility. Alipay’s platform proves that AI agents can drive commerce. We need to show that decentralized agents can do it better, with user sovereignty, lower fees, and transparent logic.
Takeaway
Yields don’t lie, but centralized platforms do cover their tracks. The next 12 months will determine whether AI agents become a commodity controlled by Big Tech or a public utility owned by users.
Watch for these signals: adoption of open agent protocols (like elizaOS or Autonolas) by actual merchants, not just token holders. Track the number of real-world transactions executed by on-chain agents. Compare that to Alipay’s internal metrics.
The blocks remember. The question is whether we’ll write the blocks fast enough.
Trust the hash, not the headline.
— Jacob Thomas, Data Detective