MCP SERVER · Model Context Protocol · AI agents

Bitquery MCP Server on-chain data for your AI agents, in plain English.

Connect mcp.bitquery.ioto Claude, Cursor or any MCP client and ask Bitquery's on-chain trading dataset anything — in plain English. The agent resolves your question to an MCP tool call and gets back real rows: DEX trades, prices, balances, holders and money flow across 40+ chains. No GraphQL to hand-write.

Connect the MCP server →Talk to sales

MCPOAuth 2.140+ chains

pipeline · MCP server

“

Ask in English

no GraphQL, no schema to learn

MCP tool call

agent resolves prompt → typed tool

⛓

On-chain answer

real rows from 40+ chains

ClaudeCursorVS CodeWindsurf

Trusted by 40,000+ developers & teams like

40+

Chains supported

1PB+

Blockchain data indexed

10B+

API calls / month

99.9%

Production uptime

Works with your agent

Connect the client you already use.

Because mcp.bitquery.io speaks the standard Model Context Protocol, any MCP-compatible client connects out of the box — over a remote HTTP/SSE transport or a local stdio bridge via mcp-remote. Authorize once over OAuth, then ask on-chain questions.

Claude

Add the server in Claude Desktop, Web or Claude Code, authorize once over OAuth, then ask on-chain questions straight from chat.

Cursor

Drop the server into .cursor/mcp.json via mcp-remote and your coding agent can pull live trades, balances and holders inline.

VS Code

Connect the MCP server to VS Code's agent mode over the same hosted endpoint — no node, no indexer to run.

Windsurf

Any MCP-compatible client — Windsurf and custom agents included — speaks to mcp.bitquery.io out of the box.

The tools it exposes

A small, typed tool surface for your agent.

The server exposes a handful of typed tools — the LLM uses them to discover what's in the dataset and pull rows. A plain-English prompt maps to one tool call with typed params; no hallucinated schema, no GraphQL to hand-write.

dex_trades

token_holders

balances

transfers

money_flow

nft_trades

dex_trades

# "Show the latest WETH/USDC trades on Uniswap"

dex_trades({
  network: "eth",
  protocol: "uniswap_v3",
  pair: "WETH/USDC",
  orderBy: "Block_Time desc",
  limit: 25
})

networkSTRINGchain to query, e.g. eth, solana

protocolSTRINGDEX / market filter (optional)

pairSTRINGbase / quote token symbol

tokenSTRINGtoken address for holders & balances

orderBySTRINGsort field, e.g. Block_Time desc

limitINTmax rows returned to the agent

Forget hand-writing GraphQL for your agent

Your agent shouldn't have to learn a query language to read the chain.

Wiring an LLM to on-chain data usually means teaching it a schema, writing GraphQL by hand, running an indexer and parsing raw logs — then babysitting it when the agent hallucinates a field. The Bitquery MCP server exposes the data as tools, so the agent just asks in plain English and gets clean, typed rows back.

What your agent needs

Do it yourself

Bitquery MCP server

Ask on-chain questions

Hand-write GraphQL per query

Plain English → tool call

Connect to your LLM client

Build a custom integration

Standard MCP, works out of the box

Trades, prices & balances

Run indexers and decoders

Tools over indexed data

Multi-chain in one place

A pipeline per chain

40+ chains, one endpoint

Trustworthy results

Hope the agent guesses fields

Typed tools, no hallucinated schema

Secure access

Manage keys in prompts

Hosted endpoint with OAuth 2.1

Build it this weekend

The agents people ship on the MCP server.

Each one is the same MCP server plugged into a different client — connect once, then let the agent ask on-chain questions for you.

Claude · Cursor

Build an AI trading agent

Let an agent pull live DEX trades, OHLC and new-token launches, then reason over them — "which Base tokens crossed $10M market cap in the last 24h?"

dex_trades({ network: "base", orderBy: "volume desc" })

Read the docs →

Investigation

On-chain research copilot

Trace money flow, rank token holders and pull wallet history in plain English — the audit and investigation workflow without GraphQL.

money_flow({ address: "0x28c6…", direction: "in" })

Read the docs →

Setup · 60s

Plug into Claude or Cursor

Add the server to Claude Code or your .cursor/mcp.json, authorize once over OAuth, then ask on-chain questions straight from chat.

claude mcp add bitquery -- npx -y mcp-remote https://mcp.bitquery.io/mcp

Read the docs →

"We did a thorough search of the market for the best onchain data. Bitquery came out on top — and now powers all live prices across Nansen. We don't think of them as a vendor. They're a partner."

Alexander Karsten

Nansen

Plain-English prompts → typed MCP tool calls

40+

Chains your agent can query in one call

Any

MCP client — Claude, Cursor, Windsurf & more

Bitquery does the hard work of parsing blockchain transaction data into a usable form so that we don't have to. We use their interface to diagnose issues with complex transactions and their analytics as a starting point for our own.

Alex Knaggs

0x Protocol

They proved they had the technology to deliver sophisticated data solutions. We extended our support through the Binance X fellowship — building an open-source library of visualization widgets on their blockchain data.

Flora Sun

Director, Binance X

The complex raw data is available at different levels of detail and from different viewpoints — whether we need simple aggregated transfers or parameters for failed contract calls. The support is responsive, friendly and quick.

Jan Dreske

Backend Developer, Blockpit

Partnering with Bitquery has been highly cost-effective — leveraging their established infrastructure rather than building our own let us rapidly expand our blockchain support and reach a much broader segment of on-chain users.

Nick Christie

Co-Founder, Syla

Bitquery's products are very intuitive and easy to use. We currently use their products to obtain DEX-related trading and liquidity information, which saves us the manpower and tedious technical details required to develop our own system. Their excellent technical team deserves special praise; they provide near-24/7 support and resolve issues quickly. I greatly appreciate their products and work ethic.

Data Team

Ourbit

Bitquery provides the infrastructure we rely on every day. Fast, reliable, and comprehensive across the chains that matter to our business.

Webacy

webacy.com

Pricing

Start free. Scale when you ship.

Query every blockchain on every plan — no chain is paywalled. Move to commercial when you need volume, SLAs and bulk datashares.

Developer

$0 / month

Free plan for developers or small projects.

All blockchains, all plans
10 requests / minute
2 streams for testing
GraphQL IDE access

Get started free

MCP server questions, answered.

What is the Bitquery MCP server?

It's a hosted Model Context Protocol server at mcp.bitquery.io that exposes Bitquery's onchain trading dataset as tools for AI agents. An LLM client asks a question in plain English, the server resolves it to a tool call, and the agent gets back real rows — DEX trades, prices, balances, holders and money flow.

Which clients can connect to it?

Any MCP-compatible client works out of the box — Claude Desktop and Web, Claude Code, Cursor, ChatGPT, VS Code, Windsurf and custom agents — over a remote HTTP/SSE transport or a local stdio bridge via mcp-remote.

What onchain data can the agent query?

Outlier-filtered DEX trades, OHLC candles, market cap, FDV, supply, wallet balances, token holders, transfers and money flow — the same production dataset behind Bitquery's GraphQL APIs, Kafka streams and TradingView feeds, across 40+ chains including Solana, Ethereum, BSC, Base, Arbitrum, Optimism, Polygon and Tron.

How do I set it up in Claude, Cursor or ChatGPT?

In Claude Code run claude mcp add bitquery -- npx -y mcp-remote https://mcp.bitquery.io/mcp; in Cursor add the same command to .cursor/mcp.json. The first time a tool runs, the client asks you to connect and authorize. See the MCP server docs for each client.

How is access authenticated and secured?

The hosted server uses OAuth 2.1 — you authorize your client to use your Bitquery account on first connection, and choose whether tool calls always need approval or are auto-allowed. No API keys pasted into prompts.

How fresh is the data?

Near real time — the dataset updates as blocks are processed across all supported chains, so your agent reads both live and historical onchain data from the same tools.

Give your agent on-chain data this week.

Connect mcp.bitquery.io to Claude, Cursor or any MCP client and ask on-chain questions in plain English. No GraphQL, no node, no schema to teach.

Get your API key →Read the docs

Hosted & OAuth-secured · Works with any MCP client · All 40+ chains included

Bitquery MCP Server on-chain data for your AI agents, in plain English.

Connect the client you already use.

Claude

Cursor

VS Code

Windsurf

A small, typed tool surface for your agent.

Your agent shouldn't have to learn a query language to read the chain.

The agents people ship on the MCP server.

Build an AI trading agent

On-chain research copilot

Plug into Claude or Cursor

What teams say about our data

Start free. Scale when you ship.

MCP server questions, answered.

Give your agent on-chain data this week.