Recently, Web3 AI projects incorporating MCP concepts have attracted market attention, with notable projects like DARK in the Solana ecosystem and SKYAI in the BNB Chain ecosystem. MCP stands for Model Context Protocol, introduced by Anthropic on November 25, 2024. As an open protocol, MCP standardizes how external data and applications interact with and provide context to large language models (LLM). The industry often compares MCP to a USB interface for AI applications, enabling AI to plug and play and access external data and tools in a versatile manner.
While MCP was initially a hot concept in the Web2 tech circle, this new AI trend has also swept into Web3. The market believes MCP can promote the development of tracks like DeFAI and GameFAI, enhance the practicality of AI Agents, and trigger a second market wave. To help readers filter projects and grasp the trend, Odaily provides an overview of existing Web3 AI projects representing MCP integration.
DARK
DARK is an experimental MCP network operating in a Trusted Execution Environment (TEE), built on Solana and has already issued its token. Its first application automatically integrates new tools into the AIMCP server under the TEE protocol, though the product is still in development, with users able to join a waitlist via email. DARK's founder is edgar, who is also a co-founder of mtndao, a DAO focused on blockchain technology and Web3 ecosystem innovation built on the Solana blockchain.
To increase community engagement and company revenue, DARK is also developing a game called DARK Games, similar to the popular 2021 game 'Dark Forest'. In the game, SPUTNIK is the first competitive AI agent, where players use DARK tokens to interact and obtain clues. SPUTNIK leverages Dark Forest MCP and Solana MCP to enable certain game functions and on-chain interactions.
- Token Contract: 8BtoThi2ZoXnF7QQK1Wjmh2JuBw9FjVvhnGMVZ2vpump
- Current Market Cap: $16 million
SkyAI
SkyAI is built on BNB Chain, aiming to use MCP to construct a blockchain-native AI infrastructure, providing a scalable and interoperable data protocol designed for Web3 AI applications. It has already issued its token. SkyAI plans to simplify AI development in blockchain environments by integrating multi-chain data access, AI agent deployment, and protocol-level utilities. Future plans include integrating advanced AI protocols like Google's A2A. However, SkyAI currently lacks a concrete product.
As the first pre-sale project supported by Four.meme platform, SkyAI was highly successful. Launched on April 17 and concluding on April 19, it raised over 83,851 BNB, valued at approximately $50 million, 167 times oversubscribed. The official announcement stated no token retention, with 80% distributed through public pre-sale and 20% allocated to liquidity, benefiting early pre-sale participants.
- Token Contract: 0x92aa03137385f18539301349dcfc9ebc923ffb10
- Current Market Cap: $40 million
Conclusion
Currently, most Web3 AI projects combining MCP protocol are still in the early development stage, with no actual products launched. However, some projects have already issued tokens. Although MCP is the latest and hottest concept in the AI Agent track, given the previous collapse of AI Agent projects, we must consider whether MCP is just another vague hype transmitted from Web2 to Web3, or if it can truly lead AI Agents towards practicality.
Web3 AI Agents have long faced two major issues. First, there is a severe disconnect between actual products and narratives. People often jokingly say that AI developers who are insignificant in Web2 can become top-tier AI developers in Web3 just by telling a story, attracting market attention. When the grand narrative is ultimately punctured by repeatedly delayed actual products, everything comes to an end.
For example, projects now claiming that with MCP, AI agents can interact with on-chain data and external tools to achieve functions like checking balances, issuing tokens, or performing DeFi arbitrage, but during the previous DeFAI boom, top projects were also telling similar stories. Investors likely haven't completely forgotten. Now they tell us that combining MCP can better achieve these functions - does this mean the previously hyped DeFAI products never actually materialized?
The second main issue is that tokens are unrelated to actual products. If they don't issue tokens, no one will pay attention. Web3 AI developers are well aware of this, so we often see token issuance and pre-sales arriving before actual products. The problem is that even after product launch, most tokens lack utility or governance connections with the actual product. With tokens deviating from products, the product's control over token prices diminishes, with track narratives and market sentiment driving price trends, while the product remains perpetually ignored.
We have already experienced a round of Web3 AI Agent disillusionment. Investors have learned their lesson and no longer easily buy into empty narratives and inflated valuations, instead seeking projects with actual product-market fit. In this light, MCP's path remains long and challenging.
