Base MCP Server

A read-only MCP (Model Context Protocol) server that lets any MCP-compatible AI agent — Claude Desktop, Claude Code, or any other MCP client — inspect wallets and contracts on Base directly through natural conversation, instead of manually looking things up on a block explorer.

This server distills five standalone tools from this collection (a transaction monitor, a contract verifier, an airdrop eligibility checker, a rug-pull risk scorer, and a token approval checker) into one always-available set of MCP tools, backed by the same well-tested Base data logic.

Tools exposed

Tool	What it does
get_wallet_balance	ETH balance of an address
get_recent_transactions	Most recent transactions for an address
get_wallet_activity	0-100 heuristic on-chain activity score
check_contract_verification	Source verification, proxy status, owner status
check_rugpull_risk	0-100 heuristic scam-pattern risk score
check_token_approvals	Active ERC-20/NFT approvals granted by a wallet

⚠️ This server is read-only

It never asks for or handles a private key, and cannot send transactions, sign anything, or revoke approvals — every tool only reads public Base blockchain data (via the Blockscout Pro API and Base's public RPC node).

Setup

1. Requirements

• Python 3.10+
• A free Blockscout Pro API key: https://dev.blockscout.com/ → Login → create a key

2. Install

git clone https://github.com/YOUR_USER/base-mcp-server.git
cd base-mcp-server
pip install -r requirements.txt

3. Configure your MCP client

Claude Desktop — edit your claude_desktop_config.json (Settings → Developer → Edit Config) and add:

{
  "mcpServers": {
    "base-tools": {
      "command": "python",
      "args": ["/absolute/path/to/base-mcp-server/server.py"],
      "env": {
        "BLOCKSCOUT_API_KEY": "your_key_here"
      }
    }
  }
}

Claude Code — from the project directory:

claude mcp add base-tools --env BLOCKSCOUT_API_KEY=your_key_here -- python /absolute/path/to/base-mcp-server/server.py

Any other MCP client — point it at python server.py with BLOCKSCOUT_API_KEY set in its environment; the server speaks standard MCP over stdio.

Restart your client after editing the config. You should see base-tools (with its 6 tools) show up as an available MCP server.

4. Try it

Once connected, just ask your agent things like:

• "What's the ETH balance of 0x...?"
• "Check this Base contract for rug-pull risk: 0x..."
• "Does this wallet have any risky unlimited token approvals?"
• "Is this contract's source code verified on Base?"
• "Show me the last 10 transactions for this address."

The agent will call the relevant tool automatically based on what you ask.

How it works

• All tools talk to Base (chainId 8453) through the free, Etherscan-compatible Blockscout Pro API, plus a couple of direct JSON-RPC calls to Base's official public node (mainnet.base.org) for on-chain reads like owner() and allowance() that are more reliable done directly than through an explorer's compatibility layer.
• Every tool degrades gracefully: if an API call fails, the tool returns a clear text message explaining that, instead of crashing the whole server or the agent's turn.
• The rug-pull risk score and activity score use the exact same heuristics as the standalone Base Rug Pull Early Warning and Base Airdrop Eligibility Checker tools — see those repos for more detail on the scoring logic.

Limitations

• Heuristic tools (check_rugpull_risk, get_wallet_activity) are not a substitute for a real security audit or official program criteria — always verify independently before acting on the results.
• Transaction and approval scanning samples up to 1,000 recent events for performance; very old activity outside that window won't be reflected.
• Depends on the free tiers of the Blockscout Pro API and Base's public RPC node; if either is rate-limited or briefly down, the affected tool will report that instead of returning data.

License

MIT — use it, modify it, fork it freely.