Blockscout MCP Server

Quick Verdict

Blockscout MCP Server is an official Blockscout integration for giving Claude, Cursor, Codex, and internal agents structured access to EVM explorer data. It covers chain discovery, addresses, ERC-20 holdings, NFTs, blocks, transactions, verified contract ABIs and source files, and read-only contract calls. As of July 21, 2026, the latest stable GitHub release is v0.16.0. The repository’s main branch already carries a 0.17 development version, but an unreleased development number is not a stable release.

Its strongest feature is a consistent tool surface across many EVM networks that use Blockscout, with response slicing, opaque pagination, progress updates, and follow-up instructions designed for limited model context. It is not a wallet, an execution engine, or an authority on economic meaning. Results depend on Blockscout PRO API responses, Chainscout metadata, and each explorer’s indexing state. Reorganizations, stale indexes, proxy contracts, token decimals, duplicate symbols, labels, and decoded-event assumptions can all change an apparently simple answer. Material findings must be checked against the intended network, block height, receipts, logs, and verified code.

Best For

• Teams building explorer assistants, wallet-support tools, contract research agents, or internal on-chain investigation workflows.
• Developers who need a shared query interface across Blockscout-supported networks and can manage chain IDs, API credits, caching, and audit logs explicitly.
• Security researchers who want ABI, verified source, address activity, and transaction details in an IDE while keeping a human verification step.
• Organizations willing to self-host the server but still use the upstream PRO API and review telemetry, licensing, credentials, and data boundaries.
• It is a poor fit for transaction signing, unattended movement of funds, deterministic real-time settlement, or investment conclusions produced without qualified review.

Key Features

• Chain discovery: get_chains_list filters networks by name, chain ID, native currency, or ecosystem. Every subsequent call should carry the verified chain_id rather than relying on an agent’s guess.
• Address and asset queries: address details, ERC-20 balances, NFT holdings, transactions, and token transfers use structured responses and opaque cursors for large result sets.
• Block and transaction tools: agents can map time to a block, inspect a block or transaction, and receive decoded parameters and token transfers. These are indexed interpretations, not consensus proofs.
• Contract inspection: the server retrieves verified ABIs and source files and can execute read-only view functions. Proxy implementation, storage layout, call context, and target block still require separate reasoning.
• Advanced API access: direct_api_call reaches raw Blockscout endpoints for chain-specific needs. Its flexibility also warrants endpoint and method allowlists, parameter validation, response caps, and tighter review.
• Context controls: standardized ToolResponse objects, field truncation, pagination, progress notifications, notes, and suggested next calls help avoid overflowing an LLM context window.
• Multiple transports: the hosted MCP endpoint is https://mcp.blockscout.com/mcp. A self-hosted server runs over stdio by default, can enable Streamable HTTP, and can add the versioned /v1 REST API with --rest.

Use Cases

A careful address investigation starts by resolving the intended network with get_chains_list, then querying transactions or transfers within a bounded time range, and finally opening the relevant transaction and verified contract. This is safer than asking a model to infer a chain from a familiar token symbol. For contract research, inspect both proxy and implementation addresses, confirm verification status and ABI, state the block being analyzed, and distinguish an event from a state transition or economic transfer.

Production deployments should expose only the required tools. Put an allowlist around direct_api_call, cap pages and response size, set timeouts and a credit budget, and test failure behavior for missing keys, exhausted credits, unsupported chains, and upstream timeouts. Never pass seed phrases, wallet private keys, exchange credentials, or unrelated secrets. Logs can retain caller, tool, chain ID, latency, result size, and error category, but should redact API keys and sensitive business parameters.

On-chain interpretation needs an explicit review boundary. An explorer label is not proof of identity. An ERC-20 transfer event is not always equivalent to beneficial ownership. A transaction’s success status does not prove the intended business outcome. A current balance can change after a reorganization or an indexing delay. Sanctions, fraud, ownership, accounting, tax, and investment assessments require qualified reviewers and independent evidence; the model’s narrative is only a research aid.

Pricing

Most hosted data tools in v0.16.0 require a Blockscout PRO API key prefixed with proapi_. The Developer Portal can issue a free-tier key without a credit card, according to the official repository. Free does not mean unlimited: calls consume a daily credit allowance, exhausted credits can produce 402 Payment Required, and sustained workloads may need additional credits or a commercial arrangement. Claude’s managed Connector uses a shared access mechanism and requires an eligible paid Claude plan; its restrictions should not be treated as the same entitlement as a caller-owned key.

The source is visible and can be self-hosted, but the project is not open source under an OSI-approved license. The Blockscout Software Licence effective May 15, 2026 is a custom source-available license. Its grant is described as temporary, revocable, non-transferable, and non-sublicensable; it requires attribution and branding, restricts derivative works to internal use, and prohibits paid, hosted, managed, or other monetized third-party use without a Commercial Licence. Review the current repository LICENSE and obtain legal advice before modifying, redistributing, embedding, or offering the server commercially. A public GitHub repository does not imply MIT, Apache, or unrestricted open-source rights.

Pros

• One tool catalog covers network discovery, addresses, fungible tokens, NFTs, blocks, transactions, ABIs, verified source, and read-only contract calls.
• Hosted MCP, local stdio, Streamable HTTP, and optional REST interfaces support several deployment patterns.
• Structured pagination, response slicing, and follow-up instructions are more agent-friendly than returning entire explorer payloads.
• A no-credit-card PRO free tier lowers the cost of a constrained proof of concept.
• The official repository publishes releases, container guidance, API documentation, and explicit authentication behavior.

Cons

• The custom Blockscout Software Licence materially limits commercial services, derivative works, redistribution, and branding; it is not a standard open-source dependency.
• Most data access requires a valid PRO key and available credits, so the hosted endpoint is not an unlimited public data source.
• Chainscout metadata, a Blockscout index, and the chain execution layer can disagree temporarily in coverage, freshness, or height.
• direct_api_call expands the misuse surface and can let an agent select an incorrect endpoint, method, network, or time range.
• Blockchain data is easy to over-interpret. Labels, proxy contracts, token symbols, decimals, logs, and transaction status require context and corroboration.
• Community-hosted instances collect anonymous telemetry by default, including tool arguments, version, and a one-way key fingerprint, unless the operator disables it.

Alternatives

FAQ

What is the current stable Blockscout MCP Server version?

As of July 21, 2026, the latest stable GitHub release is v0.16.0. A 0.17 development version on the main branch should not be represented or deployed as a stable release without a corresponding release.

Is Blockscout MCP Server open source?

No. Its source is available, but the custom Blockscout Software Licence is not an OSI-approved open-source license. It imposes restrictions on commercial services, derivatives, redistribution, attribution, transfer, and sublicensing.

Does the free tier require a credit card?

The official repository says the Developer Portal free tier does not require one. It still has a credit allowance. Confirm current limits, top-up options, and commercial terms in the Portal before estimating production capacity.

How does the hosted MCP endpoint authenticate?

Connect to https://mcp.blockscout.com/mcp and send the proapi_ key in the Blockscout-MCP-Pro-Api-Key header. Do not use a wallet private key, commit the PRO key, or bake it into a distributed client binary.

Can an answer be treated as final on-chain evidence?

Not by itself. Verify the chain ID, block height, index freshness, transaction receipt, events, contract implementation, token decimals, and relevant source code. Reorganizations and interpretation errors can invalidate a plausible summary.

Can it sign or submit transactions for an agent?

The core tool set is for data retrieval and read-only contract calls, and it should never receive a wallet secret. If a separate execution tool is introduced, isolate it behind simulation, limits, explicit human confirmation, least privilege, and complete auditing.

Bottom Line

Blockscout MCP Server v0.16.0 is a practical agent adapter for multi-network explorer data, but adoption should be driven by boundaries rather than tool count. Start with a free PRO key and one known chain. Test address, transaction, and contract queries alongside credit exhaustion, wrong-network input, indexing lag, pagination, and reorganization handling. Before production, pin a release, review the custom source-available license, isolate keys, assess or disable telemetry, constrain direct_api_call, and keep independent on-chain verification and human judgment in the workflow.