Eth Labels

API

Run it locally or serve it yourself

Ethereum

View labels here

Base

View labels here

Arbitrum

View labels here

Optimism

View labels here

Binance Smart Chain

View labels here

Gnosis Chain

View labels here

Celo

View labels here

MCP Server

Look up labels for Ethereum and EVM addresses from Claude, Cursor, Windsurf, VS Code, and other MCP-compatible clients.

169k+ labeled addresses and tokens across EVM chains. No clone. No build. Just run.

Requires: Node.js 18+

Setup

Pick your AI client and follow the instructions. Every option uses the eth-labels-mcp npm package — no need to clone this repo.

claude mcp add eth-labels-mcp -- npx -y eth-labels-mcp

On native Windows (not WSL), use:

claude mcp add eth-labels-mcp -- cmd /c npx -y eth-labels-mcp

Edit your config file:

{
  "mcpServers": {
    "eth-labels": {
      "command": "npx",
      "args": ["-y", "eth-labels-mcp"]
    }
  }
}

Restart Claude Desktop after saving.

Add to .cursor/mcp.json in your project root (or open Settings → Features → MCP Servers → Add):

{
  "mcpServers": {
    "eth-labels": {
      "command": "npx",
      "args": ["-y", "eth-labels-mcp"]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "eth-labels": {
      "command": "npx",
      "args": ["-y", "eth-labels-mcp"]
    }
  }
}

Add to .vscode/mcp.json in your project:

{
  "servers": {
    "eth-labels": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "eth-labels-mcp"]
    }
  }
}

Open Cline → MCP Servers → Configure, then add:

{
  "mcpServers": {
    "eth-labels": {
      "command": "npx",
      "args": ["-y", "eth-labels-mcp"]
    }
  }
}

Verify it works

Once configured, ask your AI:

Search for "uniswap" addresses

If it returns labeled Uniswap addresses — you're in. 🎉

Tools

Example

Ask your AI: "Who owns 0xa6baaed2053058a3c8f11e0c7a9716304454b09e?"

Q & A

• Where does this data come from?
  • This data is already organized by the kind folks at Etherscan. Unfortunately that data is not accessible for researchers, so we've copied the data out and into a more shareable format here.

Star History

Development

To use this API locally, start it like this:

bun run dev:api

Documentation for the API is available via swagger at http://localhost:3000/swagger

Automated Scraping

This project includes an automated scraper that logs into Etherscan and pulls label data. No manual cookie copy-paste required!

Setup

1. Create a .env file in the project root (see .env.example):

ETHERSCAN_USERNAME=your_etherscan_username
ETHERSCAN_PASSWORD=your_etherscan_password

1. Start Chrome with remote debugging (one of the following):

Option A: Clawdbot Managed Browser (recommended if you use Clawdbot)

• Clawdbot runs a managed Chrome instance at ws://127.0.0.1:18800
• The scraper will automatically detect and use it

Option B: Manual Chrome with DevTools

# macOS
/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --remote-debugging-port=9222

# Linux
google-chrome --remote-debugging-port=9222

Option C: Manual Cookie (fallback) If you can't run Chrome with remote debugging, you can manually extract cookies:

1. Log into Etherscan in your browser

2. Open DevTools → Application → Cookies → https://etherscan.io

3. Copy all cookies as a single string: name1=value1; name2=value2; ...

4. Add to .env: ETHERSCAN_COOKIE=your_cookie_string_here

5. Run the scraper:

bun run pull

How it works

The scraper uses Chrome DevTools Protocol (CDP) to connect to an already-running Chrome instance:

1. Detects Chrome running at port 18800 (Clawdbot) or 9222 (standard DevTools)
2. Connects via CDP (not launching a new browser — avoids Cloudflare automation detection!)
3. Opens a new tab and navigates to etherscan.io/login
4. Fills in credentials from environment variables
5. Waits for you to solve the CAPTCHA (if present)
6. Extracts cookies using CDP (including httpOnly cookies)
7. Closes only the login tab (browser stays running)
8. Uses those cookies for all subsequent requests

Why CDP? Connecting to a real Chrome instance avoids Cloudflare's automation detection. Unlike Puppeteer's launch(), which gets blocked, connecting to an existing browser looks like a normal browsing session.