🟦 BaseQL MCP Server 🟦

all Base chain data at your agent's fingertips 🔥

about

the SQL API from Coinbase Developer Platform (CDP) indexes all Base chain data across blocks, events, tranfers, transactions, and encoded logs, and lets you (or your agent) submit SQL queries against that rich dataset

this is incredibly powerful because you don't need blockchain data infrastructure like indexers anymore; you simply call SQL API and get hyper-fresh data (<500ms to tip of chain) at blazing-fast latencies (<200ms)

BUT agents need some help to use this powerful tool

the BaseQL MCP Server gives an agent all the info and tools it needs to call SQL API independently, enabling the user to fetch relevant Base chain data using natural language 🎉

what does BaseQL MCP Server include?

agents using BaseQL get:

• common Base contract addresses (USDC, WETH, AERO, other common ERC20s, popular NFTs, etc.)

• accurate SQL API schema context (correct column names, common mistakes)

• pre-built query templates (whale tracking, NFT analytics, gas analysis)

• ENS/Basename resolution (forward and reverse lookup)

• direct SQL execution via CDP SQL API

• built-in best practices (time filtering, performance optimization)

how agents use this

here's an example of what this unlocks for agents:

quickstart

installation

use with Claude Desktop

edit your Claude Desktop config:

Mac: ~/Library/Application Support/Claude/claude_desktop_config.json

restart Claude Desktop and ask: "what's the USDC contract address on Base?"

use with ChatGPT

note: with ChatGPT you need to input your MCP Server's URL, so you'll need to deploy it first (e.g. to Vercel)

1. enable developer mode in ChatGPT:

   • open ChatGPT settings → apps & connectors → advanced settings

   • enable developer mode (you should also have this enabled in your soul)

2. add MCP server:

   • now go back to apps & connectors

   • click create top-right

   • fill in the MCP server details:

     • name: BaseQL

     • URL: your deployed URL e.g. https://your-baseql-mcp.vercel.app

     • description: Base data powerrrrr

   • click save

3. test:

   • start new chat in ChatGPT

   • ask something like: "how many USDC transfers happened on Base in the last hour?"

   • chatGPT uses your BaseQL MCP to fetch the data & respond

local development

test via HTTP:

tools included

1. get_schema

get complete Base SQL table schemas with correct column names and best practices

example:

returns:

• all table schemas (base.transfers, base.transactions, base.events, base.blocks)

• critical rules (always filter by time!)

• common mistakes (column name corrections)

2. get_contract

get Base contract addresses by symbol

example:

supports:

• 46 tokens (USDC, WETH, AERO, DEGEN, TYBG, JESSE, etc.)

• 8 NFT collections (Basenames, Based Fellas, Base Punks, Base Gods, etc.)

• Infrastructure (Bridges, EAS, Coinbase Verifications)

• Base mainnet only

3. get_query_template

get pre-built SQL query templates

example:

available templates:

• whale_transfers - large token movements

• trending_tokens - tokens with sudden activity

• wallet_activity - complete wallet history

• gas_analysis - gas spending patterns

• nft_mints - NFT minting activity

• coinbase_verified_users - addresses associated with Coinbase accounts (i.e. that have KYC'd at some point)

• basename_registrations - new .base.eth names

• and a bunch more

4. run_sql_query

execute actual SQL queries against Base blockchain

example:

returns: realtime Base data (<500ms from tip of chain)

5. resolve_name

convert ENS/Basename to address

example:

6. get_name_for_address

reverse ENS lookup (address → name)

7. get_names_for_addresses

batch reverse lookup for multiple addresses

use cases for BaseQL MCP Server

DeFi analytics

NFT tracking

market intel

architecture

BaseQL provides:

• context (contracts, schemas, templates)

• execution (runs queries via CDP)

• intelligence (best practices, optimization)

configuration

environment variables

get CDP API credentials from CDP Portal

modes

stdio mode (for Claude Desktop):

• uncomment stdio transport in server.ts

• uses stdin/stdout

HTTP mode (for testing/custom agents):

• comment out stdio transport

• runs on PORT (default 4000)

why BaseQL MCP Server is dope

accurate schema

learned the hard way that Base SQL tables have inconsistent column names:

• base.transfers → token_address, block_timestamp

• base.events → address, block_timestamp

• base.transactions → from_address, timestamp (NOT block_timestamp)

BaseQL documents all these gotchas

guardrails

queries without time filters can scan 100GB+ and fail; BaseQL teaches agents to always filter by time first

curated context

repo of common contract addresses

realtime data

CDP SQL API is < 500ms from tip of chain, so BaseQL queries get near-instant Base data

examples

find USDC whales

track Basename registrations

license

MIT

links

• CDP SQL API Docs

• MCP Protocol