Scale or Die 2025: Building Better Remote MCPs: Web3's Answer to Auth & Monetization

In a groundbreaking presentation at Scale or Die 2025, Albert Chen, co-founder of Elite Agents, unveiled a revolutionary approach to implementing Model Context Protocol (MCP) that promises to transform how AI tools are accessed, shared, and monetized in the Web3 space.

Summary

Albert Chen's presentation introduces Elite Agents' innovative implementation of Model Context Protocol (MCP), a unified interface for AI tool calls. Chen explains the importance of AI tool calls in extending the capabilities of large language models (LLMs) and how MCP standardizes this process across different AI vendors.

The talk highlights Elite Agents' platform, described as "Shopify for AI agents," which allows users to easily publish and integrate MCP servers and tools. Chen also introduces a new framework called "Mac and the hablis," built on Solana, which addresses the challenges of monetizing AI tools through a novel approach to authentication and payments.

Chen delves into the technical aspects of their implementation, including the use of ED25519 key pairs for AI agent identification and the incorporation of the X402 standard for native internet payments. He explains how this system enables programmable monetization and token-gated access to AI tools, potentially revolutionizing the way AI capabilities are shared and monetized in the Web3 ecosystem.

The presentation concludes with a demonstration of how developers can easily create and monetize MCP tools using Elite Agents' open-source tools, highlighting the potential for rapid innovation and monetization in the AI tool space.

Key Points:

Understanding Model Context Protocol (MCP)

Model Context Protocol, introduced by Anthropic in December, serves as a unified interface for AI tool calls. Prior to MCP, different AI vendors had their own standards for function calls, tool calls, or AI plugins. MCP standardizes this process, allowing any AI language model to use tools built with this protocol.

Chen emphasizes the importance of AI tool calls in extending the capabilities of large language models (LLMs). While LLMs excel at probabilistic tasks like summarization or code generation, they lack the ability to execute actions or recall exact details. Tool calls bring CRUD (Create, Read, Update, Delete) capabilities to AI, grounding them in real-world context and data.

Elite Agents' MCP Implementation

Elite Agents has developed a platform described as "Shopify for AI agents." This platform allows users to easily publish and integrate MCP servers and tools. Unlike previous implementations that required complex setup processes, Elite Agents' solution offers a user-friendly interface where capabilities can be added to AI agents with a simple click.

The platform enables users to imbue their AI agents with various capabilities, such as grabbing trending tokens, executing trades, or minting NFTs. This ease of integration speeds up development and accelerates adoption of AI tools across different applications.

Mac and the hablis Framework

Chen introduces the "Mac and the hablis" framework, built on Solana, which addresses the challenges of monetizing AI tools. This framework implements the X402 standard, an HTTP error code standardized in 1999 for "payment required," now realized through Web3 technology.

The framework uses ED25519 key pairs, the same used for Solana public keys, to identify AI agents. Each request to an MCP server includes a signed payload, allowing for payment gating based on on-chain state. This system enables a generic discovery of tools and micropayments, opening up new possibilities for AI tool monetization.

Programmable Monetization and Token Gating

The Mac and the hablis framework enables programmable monetization of AI tools. Payments can be directed to a PDA (Program Derived Address) on a Solana program, allowing for various monetization strategies. These include token gating, where access to MCP tools is restricted based on token ownership, and subscription models with token buyback programs to solve token accrual issues.

This Web3-native approach taps into the power of programmable money, offering flexible and innovative ways to monetize AI capabilities. It opens up possibilities for developers to create and share valuable AI tools while ensuring fair compensation for their work.

Open-Source Tools and Easy Implementation

Elite Agents has released open-source tools to facilitate the creation and deployment of MCP servers and tools. With a simple NPX tool call, developers can scaffold an MCP server, and additional CLI utilities make it easy to add new capabilities.

Chen demonstrates how a simple MCP tool can be created in just 20 lines of code, complete with monetization features. This ease of development and deployment is crucial for encouraging innovation and adoption in the AI tool ecosystem.

Facts + Figures

• MCP (Model Context Protocol) was released by Anthropic in December, providing a unified interface for AI tool calls.
• Elite Agents' platform, described as "Shopify for AI agents," was released three weeks prior to the presentation.
• The X402 standard, based on the HTTP 402 "Payment Required" error code from 1999, was announced by Coinbase approximately 10 days before the presentation.
• AI agents in the Mac and the hablis framework are identified by ED25519 key pairs, the same used for Solana public keys.
• The framework includes four major entities: the MCP client (Mac and the hablis client), the MCP server (old want server), the facilitator, and the blockchain.
• Elite Agents' open-source tools allow for MCP server scaffolding with a single NPX tool call.
• A simple MCP tool can be created and monetized in approximately 20 lines of code using Elite Agents' framework.

Top quotes

1. "MCP stands for Model Context Protocol, not to be confused with MPC, which a lot of other projects have talked about today."
2. "Essentially, AI does not have CRUD built in, and tools bring CRUD."
3. "We call it elite agents, Shopify for AI agents."
4. "What X402 is is pretty interesting, because 402 is actually an HTTP error code that was standardized in 1999, same with 401, 400, and valid request, 403, forbidden, 404, resource not found."
5. "So as early as 1999, they imagined a future where there would be native internet payments."
6. "This is exciting, because if you make something Web3 native, you then tap into the power of programmable money."

Questions Answered

What is Model Context Protocol (MCP)?

Model Context Protocol is a unified interface for AI tool calls, released by Anthropic in December. It standardizes the process of extending AI capabilities across different vendors, allowing any AI language model to use tools built with this protocol. MCP simplifies the integration of external functions and data sources with AI models, enabling more powerful and versatile AI applications.

Why are AI tool calls important?

AI tool calls are crucial because they extend the capabilities of large language models (LLMs). While LLMs excel at probabilistic tasks like summarization or code generation, they lack the ability to execute actions or recall exact details. Tool calls bring CRUD (Create, Read, Update, Delete) capabilities to AI, grounding them in real-world context and data. This helps to reduce hallucinations and make AI more practical for real-world applications.

What is Elite Agents' contribution to the MCP ecosystem?

Elite Agents has developed a platform described as "Shopify for AI agents." This platform allows users to easily publish and integrate MCP servers and tools with a simple, user-friendly interface. It enables developers to add various capabilities to AI agents with just a click, speeding up development and accelerating the adoption of AI tools across different applications.

How does the Mac and the hablis framework work?

The Mac and the hablis framework, built on Solana, uses ED25519 key pairs to identify AI agents. Each request to an MCP server includes a signed payload, allowing for payment gating based on on-chain state. This system enables generic discovery of tools and micropayments, facilitating easy monetization of AI tools. It implements the X402 standard for native internet payments, realizing a long-standing vision of seamless online transactions.

What are the monetization possibilities with Elite Agents' MCP implementation?

Elite Agents' implementation enables programmable monetization of AI tools. Payments can be directed to a PDA (Program Derived Address) on a Solana program, allowing for various strategies such as token gating and subscription models with token buyback programs. This Web3-native approach taps into programmable money, offering flexible ways to monetize AI capabilities while ensuring fair compensation for developers.