MCP Servers for Crypto, Web3 and On-Chain Data

Crypto MCP servers let Claude Code and Cursor read prices, wallets and DEX data without leaving the editor. CoinGecko covers spot prices and market caps, Ethereum and Bitcoin expose blockchain reads, DexPaprika and Token Metrics add DEX analytics and scoring, and Thirdweb reaches thousands of blockchains for contract calls. Pick by data type first — price feed, on-chain read or transactional wallet — then check the CheckMCP audit score before wiring in write access.

Why does an AI agent need live on-chain data?

An agent that only has its training data is blind to markets that move every second. Crypto prices, gas costs and liquidity shift constantly, so an assistant answering questions about a token or a wallet needs a live connection rather than a cached snapshot. Crypto MCP servers close that gap by exposing price feeds, blockchain reads and DEX analytics as tools an agent can call mid-conversation, inside Claude Code, Cursor or any MCP-compatible client, instead of asking the developer to paste numbers in by hand.

The servers in this list split into three jobs: pricing and market data from CoinGecko, raw blockchain reads from Ethereum and Bitcoin, and DEX-level analytics or scoring from DexPaprika and Token Metrics. A few, like Thirdweb, also write to chains — deploying contracts or sending transactions — which is a materially different risk profile from a read-only price lookup. Matching the server to the job, rather than installing everything at once, keeps the agent's tool list short and its blast radius small, and the CheckMCP scores below are exactly the kind of signal the How to Audit an MCP Server Before Install guide recommends checking before granting write access.

What does the CoinGecko MCP server do for crypto prices?

CoinGecko is one of the most direct crypto mcp servers for price data: CoinGecko's own API page states that, as of September 2026, the platform tracks 21,510 coins across 1,498 exchanges with a market cap of $3.032T — the kind of live aggregate a price-lookup tool needs to stay current. Inside Claude Code or Cursor, the same tool call returns the numbers a trader would otherwise open a browser tab for.

The server installs via the MCPizy CLI with mcpizy install coingecko, or directly with npx -y @coingecko/coingecko-mcp. The CheckMCP audit of coingecko/coingecko-typescript scores the repository 74/100, grade C, citing 58 stars, an Apache-2.0 licence, 4 open issues and a last commit 34 days ago. That price-feed foundation is also why CoinGecko is the base for the Crypto Price Alerts recipe, which pushes price moves into Slack instead of requiring someone to check a dashboard by hand.

How do Ethereum MCP and Bitcoin MCP read blockchain data?

Ethereum and Bitcoin both describe themselves simply as Finance MCP servers for blockchain and smart contracts, and blockchain and transaction data respectively, giving an agent read access to on-chain state without running a full node. Ethereum's own developer documentation frames the chain around accounts, transactions and smart contracts, and Bitcoin's developer documentation centers on the block chain, proof of work and standard transaction types — the concepts a read tool ultimately turns into structured data instead of raw RPC output.

The two servers install the same way: mcpizy install ethereum or npx -y ethereum-mcp for Ethereum, and mcpizy install bitcoin or npx -y bitcoin-mcp for Bitcoin. Ethereum has a published CheckMCP audit of Axom-ai/ethereum-mcp: 31/100, grade F, with 0 stars, an MIT licence, 0 open issues and no commit for 14 months — worth knowing before depending on it for anything beyond casual lookups. Bitcoin has no CheckMCP audit published yet, so there is no maintenance signal to compare against Ethereum's.

Both sit in the Finance category alongside CoinGecko, and each lists the other as an alternative on its own page. Neither excerpt documents an authentication step, which fits a read-only blockchain explorer role: querying balances, transactions or contract state rather than signing anything. Treat both as data sources for an agent's context, not as a substitute for a dedicated node when correctness matters for anything security-sensitive.

DexPaprika MCP and Token Metrics MCP: DEX data and scoring

DexPaprika is a Blockchain-category server offering, in its own words, real-time DEX analytics across 20+ blockchains — trading pairs, volumes and liquidity — a different layer from CoinGecko's exchange-style price feed. The CheckMCP audit of coinpaprika/dexpaprika-mcp scores it 86/100, grade B, the highest of the servers covered here, with 42 stars, an MIT licence, 0 open issues and a last commit 2 days ago. It installs with mcpizy install dexpaprika or directly via npx -y dexpaprika-mcp.

Token Metrics sits in the same Blockchain category and adds real-time crypto market data, AI-driven ratings, and trading signals on top of raw feeds — scoring rather than just reporting numbers. It installs with mcpizy install token-metrics or npx -y @token-metrics-ai/mcp. No CheckMCP audit is published for Token Metrics, so unlike DexPaprika there is no independent maintenance score to weigh against its feature set before an agent leans on its ratings for a decision.

For a quick side-by-side, here is how the four audited servers in this list compare on the CheckMCP scale:

DexPaprika's 2-day-old commit and 0 open issues make it the freshest repository of the group at the time of the audit, while Token Metrics simply has no audit trail yet to compare against. For a DEX-trading assistant, that gap matters more than for a passive price display, since DexPaprika's own description centers on liquidity and volume — exactly the data a trading-adjacent agent reads before acting.

What do Bankless Onchain and Armor Crypto add for transactional agents?

Two more Blockchain-category servers appear alongside DexPaprika in its list of related tools: Bankless Onchain and Armor Crypto. Bankless Onchain is described as a way to query onchain data — ERC20 tokens, transaction history, wallet balances across chains — which places it closer to Ethereum and Bitcoin's read-only role than to a trading tool, useful for an agent that needs to check a wallet's holdings or history before reporting back to a user.

Armor Crypto is described differently: it lets an agent interface with blockchains — staking, DeFi, swap, bridging across multiple networks — which are write operations, not reads. That distinction matters for the human-in-the-loop question every team building a crypto agent has to answer: a read tool like Bankless Onchain can run unattended, while a swap-or-bridge tool like Armor Crypto changes wallet balances and should not be wired into an autonomous loop without an explicit confirmation step. This directory does not publish a CheckMCP audit or install command for either server, so treat both names as a starting point for further research rather than a same-day install.

Thirdweb MCP and Alby Bitcoin: wallets and Lightning payments

Thirdweb is the only server in this list built explicitly to write, not just read: its own description is to read and write to 2,000+ blockchains, covering smart contracts, NFTs, tokens and dApp development. The CheckMCP audit of thirdweb-dev/ai scores it 65/100, grade D, with 20 stars, an Apache-2.0 licence, 6 open issues and a last commit 51 days ago — a signal worth weighing given that write access carries more risk than a price lookup. It installs differently from the others too, via pip install thirdweb-mcp, alongside the usual mcpizy install thirdweb.

Alby Bitcoin appears in this catalogue's related-tools list as a way to connect a Bitcoin Lightning wallet to AI agents for instant micropayments and wallet management. No audit, licence, star count or install command is published for it here, so it belongs on a research list rather than a same-day install — the same caution that applies to Armor Crypto above applies here: a tool that manages a wallet or moves funds needs a confirmation step, not blind autonomy, and nothing in this article constitutes financial advice on whether to use one.

Between Thirdweb's broad multi-chain write access and Alby's Lightning-specific scope, the choice comes down to whether the agent needs to touch many chains generically or handle Bitcoin Lightning payments specifically. Either way, a lower or unpublished audit score is a reason to gate the tool behind a human approval step rather than a reason to skip it outright, especially once the tool can move funds instead of only reading them.

FAQ

Do these crypto MCP servers require an API key?

CoinGecko's configuration example lists a COINGECKO_API_KEY environment variable, but the page excerpt does not state whether a key is mandatory for basic price lookups. Ethereum, Bitcoin, DexPaprika, Token Metrics and Thirdweb show no authentication step in the excerpts covered here, which fits their read-focused descriptions; check each project's own repository before assuming a free tier covers production use.

Which of these is safest to install first?

Read-only servers with a published audit and a recent commit — DexPaprika at 86/100 with a last commit 2 days ago, or CoinGecko at 74/100 — carry less risk than a write-capable server like Thirdweb, scored 65/100 with a last commit 51 days ago, or Ethereum's own audit at 31/100 with no commit for 14 months. Starting with a price or data feed before adding a transaction-capable tool keeps the blast radius small while the agent's behaviour is still being tested.

Can an agent trade or move funds through these servers?

Thirdweb and, per its own description, Armor Crypto both expose write operations — contract calls, swaps, staking or bridging — while CoinGecko, Ethereum, Bitcoin, DexPaprika, Token Metrics and Bankless Onchain are described as data or read tools. None of this is financial advice, and any server that can sign a transaction or move a wallet balance should sit behind an explicit human confirmation step rather than run unattended, regardless of its audit score.