• Overview

    • Overview

  • Blockchain

    • Blockchain
  • BlocksTransactionsTop accountsContracts

  • Validators

    • Validators

  • API

    • API

  • Assets

    • Assets
  • TokensNFTsHybrid Tokens
monad

Monad agent kit

Interact with your
agent via MCP

mainnet

mainnet

Latest Block

‌
‌

exploreme

Network

Home

Socials

WebsiteTwitterDiscord

Help

Help centerBug report
Indexing the blocks
Evm0%...
Backend: —Frontend: v4.6.0-866e77fa

Powered by

STAKEME
Beta

This section is currently in development and may change.

Home preview background

Monad Agent Kit

MCP69 tools · MonadGitHub

MCP-based toolkit for interacting with the Monad blockchain through AI agents or terminal. Send transactions, inspect on-chain data, deploy & verify contracts — all from a single workspace.

Two ways to use

Subscription

Free

Connect MCP to Claude Code, Cursor, or Codex and use your existing subscription. No API key needed.

  • Claude Code auto-detects .mcp.json
  • Works with Claude Pro/Max or Cursor Pro
Claude CodeClaude Code setupCursorCursor setupCodexCodex setup

AI SDK

Developers

Build programmatic agents via Vercel AI SDK with Claude or OpenAI. Full control over agent behavior.

  • Vercel AI SDK integration
  • Claude and OpenAI support
  • Ready-to-run demos included
AI SDKAI SDK docs

Quick Start

Claude CodeClaude Code auto-detects .mcp.json and connects to Monad. Use built-in skills:

Architecture

The AI agent sees only your wallet address and tool results — never the private key. Signing always happens locally before the transaction is broadcast.

sign-tx (local)

Signs tx locallyKey in .env or keystore
unix

MCP Server (remote)

prepare_* transactionsbroadcast signed txquery blocks & tokensread contracts & verify

AI Agent (ClaudeClaude/GPTGPT/ local)

✓ Sees: wallet address, tool results✗ Never sees: private key

You (chat or code)

Security

Key Principle

The private key never leaves your machine. MCP prepares unsigned transactions, signing happens locally, then the signed transaction is broadcast back through MCP.

Simple Mode

Default

Private key stored in .env, protected by guard hooks that block the AI agent from reading it. Tests cover 26+ attack vectors.

Secure Mode

Maximum security

Private key encrypted in keystore, decrypted only in a separate daemon process. The AI agent physically cannot access the key — it only receives the signed hash.

Docker Isolation

Supply chain protection

Run all npm installs and demos inside a container. Node modules are isolated — compromised packages can't reach your host filesystem or keys.

MCP Tools

69 tools total

The Monad MCP server at https://api.monad.exploreme.pro/mcp provides tools across 6 categories.

Transactions
  • prepare_native_transfer
  • prepare_erc20_transfer
  • prepare_transaction
  • broadcast_signed_raw_transaction
  • wait_for_transaction
Balances
  • get_balance
  • get_token_balance
Blocks
  • list_evm_blocks
  • get_evm_block_by_height
Contracts
  • read_evm_contract
  • verify_evm_contract_standard_json
Tokens
  • list_erc20_tokens
  • get_erc20_token_by_address
Explorer
  • explorer_search
  • get_account_by_address