Back to MCP Directory
Starknet
Model Context Protocol moderate risk

Starknet

This MCP server provides AI agents with comprehensive access to the Starknet blockchain, enabling them to query data, manage wallets, and interact with smart contracts.

Connections & Capabilities

Connects To

GitHub

Capabilities

readwriteexec

Quickstart

Install

npx -y @smithery/cli install @mcpdotdirect/starknet-mcp-server --client claude

Config

{
  "mcpServers": {
    "starknet-mcp-server": {
      "command": "npx",
      "args": [
        "@mcpdotdirect/starknet-mcp-server"
      ]
    },
    "starknet-mcp-http": {
      "command": "npx",
      "args": [
        "@mcpdotdirect/starknet-mcp-server",
        "http"
      ]
    }
  }
}

Exposed MCP Tools (4)

safe
get_starknet_eth_balance

Retrieves the ETH balance for a given Starknet address or Starknet ID.

Read-only operation, no side effects.

safe
resolve_starknet_address

Resolves a Starknet ID to its corresponding Starknet address.

Read-only operation, no side effects.

moderate
call_starknet_contract

Calls a function on a Starknet smart contract.

Can perform state-changing operations depending on the contract function called.

safe
get_supported_starknet_networks

Retrieves a list of supported Starknet networks.

Read-only operation, no side effects.

Safety Assessment

This server offers powerful Starknet interaction capabilities, but requires careful configuration and monitoring. Read-only operations are relatively safe, while write operations demand stringent security practices to prevent unintended actions or exploits.

  • Supports read and write operations on the Starknet blockchain.
  • Includes Starknet ID resolution for enhanced usability.
  • Supports multiple networks (Mainnet, Sepolia).
  • Designed to be used with AI assistants, providing a controlled interface.
  • Requires careful management of private keys for write operations.
  • Potential for unintended consequences if AI agents are not properly configured.
  • Interaction with smart contracts carries inherent risks.
  • No explicit sandboxing is mentioned in the documentation.