CryptoAPIs MCP Servers Are Live for AI Agents

Imagine asking your AI assistant: "Has this wallet received any transactions in the last 24 hours?" — and getting an accurate, real-time answer in seconds, with zero integration code.

That works today. Here's how.

The problem with AI agents and blockchain data

If you've tried to give an AI agent access to live on-chain data, you know the friction. You write a custom tool, wrap your API calls, handle chain-specific quirks, manage authentication, and by the time your agent can query a wallet balance, check a transaction, and estimate gas, you're maintaining a sprawling set of connectors.

Model Context Protocol (MCP) — the open standard now adopted by Anthropic, OpenAI, Google, and Microsoft — eliminates that. It's the layer that lets AI models discover and call external tools without custom integration work. CryptoAPIs now has 14 purpose-built MCP servers that plug directly into that standard.

Getting started in 60 seconds

Every server runs with a single npx command. No cloning, no building. Add this to your Claude Desktop, Cursor, or Windsurf config:

{
   "mcpServers": {
       "cryptoapis": {
           "headers": {
               "x-api-key": "YOUR_API_KEY"
           },
           "type": "http",
           "url": "https://ai.cryptoapis.io/mcp"
       }
   }
}

Your agent can now answer questions like:

• "What's the current ETH balance of this address?"
• "Show me all token transfers for this wallet in the last 7 days"
• "What's the current gas fee for a standard EVM transfer?"
• "Has this smart contract been called in the last hour?"

If you'd rather skip local setup entirely, connect directly to our remote endpoint:

{
   "mcpServers": {
       "cryptoapis": {
           "type": "http",
           "url": "https://ai.cryptoapis.io/mcp",
           "headers": {
               "x-api-key": "YOUR_API_KEY"
           }
       }
   }
}

One endpoint. All 50 tools. All 14 server packages. Works with any MCP-compatible client.

The 14 servers — organized by what you're actually building

Rather than listing them alphabetically, here's how they map to real workflows:

Querying on-chain state

• mcp-address-latest — current balances, nonces, token holdings, and recent transactions across EVM, UTXO, Solana, XRP, and Kaspa
• mcp-address-history — sync addresses for continuous tracking and retrieve the complete transaction and token transfer history
• mcp-block-data — blocks by height or hash, latest mined block, transactions within a block
• mcp-transactions-data — full transaction details, internal transactions, token transfers, and logs

Building and executing transactions

• mcp-blockchain-fees — fee recommendations, EIP-1559 estimates, gas estimation for transfers and contract interactions
• mcp-simulate — dry-run EVM transactions before broadcasting. See exact outcomes, gas usage, and state changes before spending anything
• mcp-prepare-transactions — build unsigned EVM transactions (native coin, ERC-20, ERC-721) ready for signing
• mcp-signer — sign locally using your private key. No API calls made, keys never leave your machine
• mcp-broadcast — submit signed raw transactions to the network

Market and contract data

• mcp-market-data — real-time asset prices and exchange rates between any pair
• mcp-contracts — smart contract ABIs and token data from EVM and Solana

Wallet and event management

• mcp-hd-wallet — HD wallet management via xPub: sync, derive addresses, retrieve balances and transaction history across EVM, UTXO, and XRP
• mcp-blockchain-events — create webhook subscriptions for on-chain events: confirmed transactions, new blocks, token transfers, custom confirmation counts
• mcp-utils — address derivation, validation, raw transaction decoding across EVM, UTXO, and XRP

Why the remote vs. local split matters

For most AI agent use cases — querying data, monitoring wallets, checking fees — the remote endpoint at ai.cryptoapis.io/mcp is the right choice. Fast to connect, no local process to manage.

For transaction signing, the local mcp-signer exists for a reason: your private keys should never travel over the network. In a production custody or trading context, you'd combine both — remote for all data queries, local signer for transaction execution. Your keys stay on your infrastructure. Your agent retains full capabilities.

If you only need a subset of tools, the remote server also supports product-scoped endpoints — for example, https://ai.cryptoapis.io/mcp/address-latest exposes only address balance tools, which means faster session initialization and a smaller context window for your AI client.

Chains supported

• EVM — Ethereum, Ethereum Classic, BSC, Polygon, Avalanche C-Chain, Tron, Arbitrum, Base, Optimism
• UTXO — Bitcoin, Bitcoin Cash, Litecoin, Dogecoin, Dash, Zcash
• Other — XRP, Solana, Kaspa
• Mainnet and testnet support available.

Full documentation

Complete install instructions, configuration examples for every MCP client, the full tool reference, and HTTP transport setup are in the official docs:
developers.cryptoapis.io/v-2.2024-12-12-175/RESTapis/ai-compatibility/mcp-servers

All 14 servers are open source, published on npm under @cryptoapis-io, and compatible with Claude, Cursor, Windsurf, n8n, and any MCP-compatible host.
Get your API key at cryptoapis.io and let us know what you're building.