GCP Web3 Datasets MCP Server (Unofficial)

Unofficial MCP server for querying Google's Web3 public datasets on BigQuery. 17 blockchain chains, 23 tools, built-in cost controls.

Quick Start

npm install -g unofficialmcp-gcp-web3-public-datasets

Or use directly with npx:

npx unofficialmcp-gcp-web3-public-datasets

Add to Claude Desktop

Add to your Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "gcp-web3-datasets": {
      "command": "npx",
      "args": ["unofficialmcp-gcp-web3-public-datasets"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/your/bigquery-key.json",
        "GCP_PROJECT_ID": "your-project-id"
      }
    }
  }
}

Optional cost control env vars:

MAX_COST_PER_QUERY_USD=1.0    (default: 1.0)
HARD_COST_CAP_USD=5.0          (default: 5.0)
MAX_ROWS=10000                  (default: 10000)
QUERY_TIMEOUT_SECONDS=30        (default: 30)

Supported Chains

| Family | Chains | |--------|--------| | EVM | Ethereum, Polygon, Ethereum Classic | | UTXO | Bitcoin, Bitcoin Cash, Litecoin, Dogecoin, Dash, Zcash | | Solana | Solana | | Beacon | Ethereum Beacon Chain | | Custom | Band, IoTeX, Tezos, Theta, Zilliqa, MultiversX |

Tools (23)

| Category | Tool | Description | |----------|------|-------------| | Discovery | list_chains | List available blockchain datasets | | Discovery | get_schema | Get table and column details | | Discovery | get_cost_estimate | Estimate query cost (dry-run) | | EVM | evm_transactions | List individual transactions for address | | EVM | evm_block_stats | Block stats for date range | | EVM | evm_transaction_summary | Transaction summary for address | | EVM | evm_token_transfers | ERC20/721 token transfers | | EVM | evm_top_addresses | Top addresses by metric | | EVM | evm_contract_info | Smart contract details | | EVM | evm_eth_balance | Native token balance | | EVM | evm_token_info | Token symbol/name/decimals | | UTXO | utxo_transactions | List individual transactions for address | | UTXO | utxo_block_stats | UTXO chain block stats | | UTXO | utxo_transaction_summary | UTXO transaction stats | | UTXO | utxo_address_activity | Address inputs + outputs | | Solana | sol_transactions | List transactions for account | | Solana | sol_block_stats | Solana block stats | | Solana | sol_token_transfers | SPL token transfers | | Solana | sol_transaction_summary | Transaction by signature | | Beacon | beacon_validator_info | Validator status/balance | | Beacon | beacon_block_stats | Block proposal stats | | Query | build_and_run_query | Structured plan to SQL | | Query | run_raw_sql | Raw SQL with cost gate |

Cost Controls

Every query goes through a dry-run cost estimate before execution:

• Per-query limit: $1.00 (configurable)
• Hard cap: $5.00 (configurable)
• Max rows: 10,000
• Timeout: 30 seconds
• Read-only: Only SELECT queries allowed

GCP Setup

Prerequisites

• Node.js 18+
• A Google Cloud account (free tier works)

Step 1: Create a GCP Project

1. Go to Google Cloud Console
2. Click Select a project > New Project
3. Name it anything (e.g. web3-datasets) and click Create
4. Note your Project ID (e.g. web3-datasets-12345)

Step 2: Enable BigQuery API

1. In your new project, go to APIs & Services > Enable APIs
2. Search for BigQuery API and click Enable

Step 3: Create a Service Account

1. Go to IAM & Admin > Service Accounts
2. Click Create Service Account
3. Name it bigquery-reader
4. Add these two roles:
   • BigQuery Data Viewer (reads public datasets)
   • BigQuery Job User (runs queries)
5. Click Done

Step 4: Download the Key

1. Click on your new service account
2. Go to the Keys tab
3. Click Add Key > Create new key > JSON
4. Save the downloaded JSON file somewhere safe (e.g. ~/bigquery-key.json)

Cost Expectations

The public blockchain datasets are free to access. You only pay for query compute (bytes scanned):

• BigQuery pricing: $6.25 per TB scanned
• Most single queries cost $0.00 - $0.01
• The full test suite costs ~$0.34
• The built-in cost gate rejects any query over $1.00 before it runs

Architecture

Agent (Claude) → MCP Tool Call → tools → api-client → BigQuery SDK → Google BigQuery
                                   ↓
                             query-builder (structured plan → SQL)
                             sql-sanitizer (BigQuery dialect fixes)
                             config (chain registry, cost limits)

Two Query Modes

1. Structured Plan (build_and_run_query): Agent provides a structured query plan (columns, filters, aggregations). The server generates correct BigQuery SQL deterministically.

2. Raw SQL (run_raw_sql): For complex queries (CTEs, JOINs, window functions). SQL is sanitized for BigQuery dialect, validated read-only, then dry-run cost checked before execution.

License

MIT