Use BlockReq with AI Assistants

BlockReq publishes its docs in forms that AI assistants and coding agents can read directly, and runs a public MCP server. Everything below is generated from the same catalogs and templates as these pages, so it stays in sync with them.

Machine-readable docs

Resource	What it holds
llms.txt	An index of every docs page, grouped by topic, following the llmstxt.org format
llms-full.txt	The core docs in one Markdown file, with a network capability table and every JSON-RPC method once
Markdown pages	Every page is also available as Markdown: add `index.md` to its URL, for example `https://blockreq.com/docs/build/api-reference/ethereum/eth_getBalance/index.md`
networks.json	Every published network with its host, public URL, private URL pattern, WSS, and archive and proof/debug/trace flags
OpenRPC	One OpenRPC document per EVM network, listing exactly the methods documented for it

Point a general-purpose agent at https://blockreq.com/docs/llms.txt and let it follow the links it needs.

The blockreq-rpc Agent Skill

The blockreq-rpc skill teaches a coding agent how to call BlockReq: endpoint URLs, how to pick the network host, which methods need archive data or are limited to proof/debug/trace networks, eth_subscribe over WSS, Request Counting and plan limits, short viem, ethers, curl, and web3.py examples, and common pitfalls.

Browse it: SKILL.md, with networks and methods references.
Download it: blockreq-rpc.zip.
Install it from GitHub: blockreq/skills is a Claude Code plugin marketplace that syncs this skill from the ZIP above every day.

Claude Code

Install the plugin from the blockreq/skills marketplace:


/plugin marketplace add blockreq/skills
/plugin install blockreq-rpc@blockreq

Or copy the skill in yourself. For your user, so every project can use it:


mkdir -p ~/.claude/skills && cd ~/.claude/skills
curl -fsSLO https://blockreq.com/docs/agent-skills/blockreq-rpc.zip
unzip -o blockreq-rpc.zip && rm blockreq-rpc.zip

For one project, unzip it into that project's .claude/skills/ folder so it is committed with the code. Either way the skill lives in a blockreq-rpc/ folder, and Claude Code loads it when a task involves BlockReq RPC.

Claude.ai

Download blockreq-rpc.zip and upload it as a custom skill in Claude.ai. Custom skills need a Pro, Max, Team, or Enterprise plan with code execution enabled; see Anthropic's Using Skills in Claude for where to upload.

Other agents

Agents that do not read Agent Skills can use https://blockreq.com/docs/llms.txt, or load SKILL.md and its reference files as instructions.

The BlockReq MCP server

https://mcp.blockreq.com/mcp is a remote MCP server over Streamable HTTP. It is public, free, and read-only, and it works without an API key or sign-in. It has fourteen tools: six for docs, the network catalog and JSON-RPC, and eight for on-chain DEX and wallet data. Connect it with your BlockReq API key as a bearer token (Authorization: Bearer <API key>) and call_rpc uses your own private endpoint instead of the shared public trial; send your own Crypto API key in the X-Data-Api-Key header and the data tools use your plan's limits instead of the anonymous limit.

Tool	What it does
`list_chains`	Lists every published network with its host, public URL, private URL pattern, WSS, and archive and proof/debug/trace flags
`get_chain`	Returns one network's URLs, flags, API reference page, and documented JSON-RPC methods
`search_docs`	Searches these docs, in English or Chinese, and returns titles, URLs, and snippets
`get_doc_page`	Returns one docs page as Markdown
`get_method_reference`	Returns a method's parameters, result, example, and endpoints on one EVM network, from its OpenRPC document
`call_rpc`	Calls a read-only method on an EVM network's public trial endpoint, or on your own endpoint when you connect with an API key or a signed-in account: `eth_blockNumber`, `eth_chainId`, `eth_getBalance`, `eth_call`, `eth_getLogs`, `eth_getTransactionByHash`, `eth_getTransactionReceipt`, `eth_getCode`, `eth_getStorageAt`, and `eth_getBlockByNumber`
`search_tokens`	Finds tokens by symbol, name, or address on a chain, each with its deepest pool, USD price, liquidity, and 24h volume
`get_token`	One token's USD price with the block and time it was read at, 24h change and volume, the pool the price comes from, a low-liquidity flag, the price's age, `stale` (a stale price must not be shown as current) and how far the chain's indexing has got
`get_token_prices`	Current USD prices of up to 50 tokens in one call, across chains, each with `price_time`, `age_seconds` and `stale`
`get_token_candles`	A token's USD candles (open, high, low, close, volume, trades), oldest first
`list_token_pools`	The pools that hold a token, deepest first
`list_pool_trades`	A pool's recent trades, newest first, with block number, time, transaction hash, trader, USD value, and which of token0 or token1 the trader bought
`get_token_holders`	A token's largest holders with balance and share of the known supply, as of a block; only tokens BlockReq tracks have a ranking
`get_account_portfolio`	An address's token holdings on one chain with USD values, including its native coin

The data tools cover Ethereum, Base, and Arbitrum, are the same data as the On-chain DEX API and the Wallet API, and are read-only. Every result carries the block and time it is as of and the exact public API request it came from, so you can repeat it with curl. Without a key they are anonymous: each call counts against the anonymous per-IP limit of the BlockReq Crypto API, shared with every other client on the same IP address, and a 429 answer says how many seconds to wait. Pass chain to search_tokens; a search across every chain can time out. Send X-Data-Api-Key: <Crypto API key> and the calls use your own Crypto API plan; a key the API rejects fails the connection with 401 instead of falling back to the anonymous limit. This header is separate from Authorization: Bearer, which stays the RPC endpoint key for call_rpc; send either, both, or neither. Both endpoints list all fourteen tools; on the account endpoint below the data tools use your own key from your account instead of a header.

Without a key, call_rpc goes through the same rate-limited public trial endpoint as https://<host>/v1/rpc/public, so it serves recent state only, without deep history, debug, or trace, and every client that shares an IP address shares that limit. With a key, it calls https://<host>/v1/rpc with your key in the X-API-Key header: the call counts against your account's Requests (the Free plan includes 3,000,000 every 30 days) and gets your plan's limits and history, and the key must belong to an endpoint that covers the network. The server never returns or logs your key. eth_getLogs takes numeric fromBlock and toBlock at most 1,000 blocks apart, or a blockHash, and responses above 256 KB are refused. The server cannot sign or send transactions. Production code should call a private endpoint with an API key instead.

Claude Code


claude mcp add --transport http blockreq https://mcp.blockreq.com/mcp \
  --header "Authorization: Bearer $BLOCKREQ_API_KEY"

Leave out --header to use the public trial. To use your own Crypto API key for the data tools, add --header "X-Data-Api-Key: $BLOCKREQ_DATA_API_KEY". Add --scope user to make it available in every project. See Connect Claude Code to tools via MCP.

Cursor

Add the server to ~/.cursor/mcp.json, or to .cursor/mcp.json in one project:


{
  "mcpServers": {
    "blockreq": {
      "url": "https://mcp.blockreq.com/mcp",
      "headers": {
        "Authorization": "Bearer ${env:BLOCKREQ_API_KEY}",
        "X-Data-Api-Key": "${env:BLOCKREQ_DATA_API_KEY}"
      }
    }
  }
}

Leave out headers to use the public trial, or keep only the header you need. Cursor may need a restart to list a new server.

See Cursor's Model Context Protocol docs.

Claude Desktop

Claude Desktop adds remote servers as custom connectors; it does not read them from claude_desktop_config.json, which is for local servers. Open Customize > Connectors, choose Add, then Add custom connector, enter the name BlockReq and the URL https://mcp.blockreq.com/mcp, and finish with Add. To use your own keys, add X-Data-Api-Key (and Authorization: Bearer <API key> for call_rpc) under Request headers, and set Authentication to No sign in. Then enable the connector for a conversation from the + button. Claude Desktop reaches the server from Anthropic's servers, so its users share one IP address and one anonymous limit; add your own Crypto API key for your plan's limits. See Anthropic's Get started with custom connectors using remote MCP.

Claude.ai and other hosted clients: connect your account

Hosted clients run on their provider's servers, so their users share a few IP addresses and one public trial limit. Add https://mcp.blockreq.com/mcp/account instead:

The client sends you to blockreq.com to sign in and approve access.
call_rpc then uses your active endpoint on each network, and every call counts against your Requests.
The eight data tools use the key of your Crypto API endpoint, so each call counts against your Crypto API plan like any call with that key. Approve "Read DEX and wallet data" on the consent page; a connection made before that option existed has to be disconnected and connected again.
You need an active endpoint for each network you call, and a Crypto API endpoint for the data tools; create them at blockreq.com/dashboard. The consent page tells you whether a free Crypto API endpoint is created for you.
Call the data tools one at a time: the Free plan allows 1 request per second with a burst of 1, and a call that is limited for 2 seconds or less is retried once for you.
The client never receives your API keys.
To disconnect an assistant, open Settings → Connected assistants in the Dashboard and choose Disconnect; it loses access within seconds and cannot refresh its sign-in.
Claude.ai: add https://mcp.blockreq.com/mcp/account as a custom connector and connect when asked; see Anthropic's Get started with custom connectors using remote MCP. To try it without an account, add https://mcp.blockreq.com/mcp instead.
Other clients that support MCP authorization (OAuth 2.1 with dynamic client registration) connect to https://mcp.blockreq.com/mcp/account the same way.

Other clients

Any client that supports remote MCP servers over Streamable HTTP: use the URL https://mcp.blockreq.com/mcp, and add the headers Authorization: Bearer <API key> and X-Data-Api-Key: <Crypto API key> if the client allows custom headers.
Clients that only start local (stdio) servers can bridge to it with npx mcp-remote https://mcp.blockreq.com/mcp.
Agents running in the browser that support WebMCP find the same tools on blockreq.com pages, which also serve the public endpoint at the same-origin https://blockreq.com/mcp.