Skip to main content
The Model Context Protocol (MCP) is an open protocol designed to allow AI agents, IDEs, and automation tools to consume, query, and analyze structured data through context-aware APIs. This server wraps Blockscout APIs and exposes blockchain data like balances, tokens, NFTs, and contract metadata via MCP so that AI agents and tools (like Claude, Cursor, or IDEs) can access and analyze it contextually. Connect it to Claude, Claude Code, Codex, or Cursor and query chains in plain English, with answers that come straight from live Blockscout data.
A single connection reaches every supported chain. Name the network in your prompt and the agent resolves its chain ID, or call get_chains_list to browse them all.

Key Features

  • Contextual blockchain data access for AI tools
  • Multi-chain support across 100+ hosted EVM chains through the Blockscout API gateway
  • Custom instructions for MCP host to use the server
  • Intelligent context optimization to conserve LLM tokens while preserving data accessibility, including smart response slicing, opaque cursor pagination, and automatic truncation of large data fields with clear indicators
  • Supports MCP progress notifications for multi-step tool operations, allowing clients to track execution status
  • Enhanced User Experience: Provides periodic progress updates for long-running API queries (e.g., fetching extensive transaction histories) when requested by the client, improving responsiveness

Use Cases

  • Enabling AI agents to analyze blockchain transactions and account balances.
  • Automating blockchain data analysis workflows with context-aware APIs.
  • Allowing IDEs to fetch contract ABIs and other relevant smart contract information.

Connection details

The server uses the Streamable HTTP transport. With your own Blockscout API key you get full rate limits, the complete tool set, and the full set of chains. Without one, the server falls back to a shared gateway key with stricter rate limits and restricted access to some chains and datasets.
Only paste your API key into the official server at https://mcp.blockscout.com/mcp.
Some clients can only send header names from a fixed list, such as Claude custom connectors. For those, send the key in an x-api-key header instead. If both headers are present, Blockscout-MCP-Pro-Api-Key wins.

Get started

The steps below connect the server with your own key. First, create a free API key: register for an account, then generate a key from your dashboard. No credit card is required. Then pick your client below and follow it end to end.
Claude on the web, Claude Desktop, and Cowork all connect through a custom connector.1. Add the connectorGo to Customize > Connectors and click Add custom connector. On Team and Enterprise plans, an organization Owner does this under Organization settings > Connectors.Enter the following details:
  • Name: Blockscout
  • URL: https://mcp.blockscout.com/mcp
  • Authentication: leave as None. A warning that the connector has no credentials is expected.
  • Request headers: select x-api-key (exactly this name) and paste your API key as the value, then click Add.
Request headers is in beta. If your dialog does not show it, install Blockscout from the Anthropic Connectors Directory instead. The directory connector uses a shared access key, which currently allows up to 5 consecutive MCP tool calls. That is enough to try the server, register for a Blockscout API account, get your personal key, and set up the custom connector above for full access.
Authentication settings cannot be edited after a connector is added. To change your key, remove the connector and add it again.
2. Run a first promptOpen a new chat and try a prompt. Approve the tool calls when Claude asks:
The agent uses get_block_number to fetch the latest block and its timestamp.

Examples

Try these in any connected client. The agent decides which tools to call:
For a longer walkthrough, see how to use the MCP to find ENV variables for rollup deployments.

Tools

The server exposes 16 tools. Your agent chooses which tool to call, so prompts don’t need to name them. The same tools are also available as versioned REST endpoints under /v1/....

Good to know

  • Read-only. The server only reads on-chain data. It cannot sign or broadcast transactions.
  • Tool calls use your API credits. The server reads data through the Blockscout API, so calls made with your key are metered against your plan. When your balance runs low, tool responses can include an advisory note. If credits run out, calls fail with a payment-required error (HTTP 402 on the REST endpoints).
  • Without a key, limits are tighter. The shared gateway key has stricter rate limits and may restrict access to certain chains or datasets. For full access, add your own free key.
  • Responses are trimmed for agents. Large results are sliced with opaque pagination cursors, and long data fields are truncated with clear indicators, so agents keep context without losing access to the data.
  • Keep tool approvals on. Data returned by any MCP server can contain untrusted text, so leave your client’s approval prompts enabled, especially when other MCP servers share the session.

Architecture and Data Flow

Every data request is validated against the supported chain list, then sent to the authenticated Blockscout API gateway with your key. How a session flows:
  1. Unlock. The agent calls __unlock_blockchain_analysis__ once to receive server reference data and a pointer to the blockscout-analysis skill.
  2. Chain selection. get_chains_list joins the chains supported by the Blockscout API with chain metadata (name, ecosystem, testnet flag, native currency) from Chainscout. Only chains present in both sources are listed. The result is cached.
  3. Data requests. Each tool call is validated against the cached chain configuration, then sent to the Blockscout API. Independent requests run concurrently. For example, get_address_info makes three at once: basic on-chain data, the first transaction, and address metadata. A failure in a secondary request is reported in the response notes instead of failing the whole call.
  4. Key handling. Your API key is sent only to the Blockscout API. ENS resolution and Chainscout never receive it, and chain discovery works without a key.

REST API data flow

The same tools are also available as REST endpoints under /v1/. These are thin wrappers that call the same tool functions as the MCP server.

Go further with agent skills

The MCP server gives your agent the tools. The blockscout-analysis skill teaches it to use them well: how to page results, handle rate limits, fall back to the REST API, and script ad-hoc analysis. The web3-dev skill teaches it to build apps, bots, and indexers directly on the Blockscout API. Install both from the agent-skills repository, or see the AI agents page for per-agent install commands.

Docs for agents

Point your agent, RAG pipeline, or coding assistant at these URLs for accurate, up-to-date Blockscout knowledge: