MCP.so
登录

概览

What is Web3 MCP?

Web3 MCP is a Model Context Protocol server implementation that wraps Ankr’s Advanced API, enabling LLMs to interact with blockchain data. It supports multiple networks including Ethereum, BSC, Polygon, and Avalanche, and is intended for developers building AI-powered blockchain applications.

How to use Web3 MCP?

Install with uv pip install -e . from the cloned repository, then set the ANKR_ENDPOINT environment variable to your Ankr RPC endpoint. Optionally set ANKR_PRIVATE_KEY for authenticated requests. Run the server via python -m web3_mcp or use the provided init_server function in Python.

Key features of Web3 MCP

  • Complete wrapper for all Ankr Advanced API endpoints
  • NFT API: metadata, holders, transfers, and ownership
  • Query API: blockchain stats, blocks, logs, and transactions
  • Token API: balances, prices, holders, and transfer history
  • Support for multiple blockchain networks

Use cases of Web3 MCP

  • Retrieve NFT metadata and ownership for a wallet address
  • Query token balances and prices across supported chains
  • Fetch transaction history and logs for addresses or contracts
  • Access blockchain statistics and block information programmatically
  • Build LLM-driven agents that interact with on-chain data

FAQ from Web3 MCP

What does Web3 MCP do?

It wraps Ankr’s Advanced API to give an MCP server access to blockchain data for NFTs, tokens, queries, and logs across multiple chains.

How do I install Web3 MCP?

Clone the repository, then run uv pip install -e . in the project directory.

What environment variables are required?

ANKR_ENDPOINT must be set to your Ankr RPC endpoint. ANKR_PRIVATE_KEY is recommended but optional for authenticated requests.

Which blockchains are supported?

The server supports Ethereum, BSC, Polygon, Avalanche, and other chains accessible via the Ankr Advanced API.

How do I run the server?

Set the environment variables and execute python -m web3_mcp from the command line, or call init_server() programmatically and then mcp.run().

标签

来自「其他」的更多内容