MCP Server for AI readers

Project name

This project, “MCP server for AI readers,” is an MCP server that provides AI with documents stored on the blockchain.

Project description.

This project is a demo application that proposes a way to provide "document location information" to AI agents.

Contract: Stores the title of a document (book) and the URL where it can be obtained. This information is recorded on the blockchain as an NFT.

Frontend: Acts as an MCP server. It receives a book title from the LLM host, queries the information stored on-chain, and returns the URL corresponding to the title.

As AI agents become more widely used, they will gather diverse information and present decision proposals. However, because they heavily rely on pre-trained LLM knowledge and information fetched from MCP servers, it becomes difficult for humans to verify the evidence behind those reports and proposals. There is also a risk of unintended bias in the data sources that agents use.

To address this, we aim to provide documents in formats that are also friendly to humans—such as books, magazines, and papers—so we can improve evidence traceability and mutual intelligibility. Just as people can freely read books at bookstores and libraries, AI should be able to freely and programmatically consult paid or free content. For this reason, this application records document metadata on the blockchain.

Technologies used

• Frontend
  • Next.js
  • Vercel MCP Server Template(https://vercel.com/docs/mcp/deploy-mcp-servers-to-vercel)
• Contract
  • Hardhat
  • OpenZeppelin (ERC‑721)

Basic architecture

The app is essentially an MCP server. When the MCP server responds with results, it references the blockchain.

Source code 

https://github.com/naokomada/mcp-server-for-ai-readers/tree/main

Deployment

Vercel(This project is an MCP server and cannot be used in a web browser.)

https://mcp-server-for-ai-readers.vercel.app/mcp

Deployed contract

• Sepolia
  • https://sepolia.etherscan.io/address/0x07BaD34F93032cfaD3432DdD1b1BdFC32DDddb7D