Model Context Protocol server for blockchain data and transaction signing — sign in a browser wallet or on your phone, with simulation and ERC-7730 clear signing before you approve
Web3-tools-mcp has 10 tools spanning blockchain operations (read-only and irreversible). Critical gaps: input schemas are not visible in the provided source code, making it impossible to verify parameter types, descriptions, or constraints. Tool descriptions exist in package.json but are absent from individual tool implementations. Two high-risk irreversible tools (sign_transaction, send_transaction) lack documented error handling, confirmation mechanisms, and recovery guidance. Naming is generally clear (verb_noun pattern: get_balance, resolve_ens_name, send_transaction), but schemas and parameter documentation are not verifiable from the code provided. This server would require inspection of the actual TypeScript tool definitions (mcp/src/tools/*.ts files) to assess schema quality properly.
Input schemas not visible in provided source code. Cannot verify parameter types, descriptions, constraints, or defaults. All 10 tools score 0 on schema dimension.
Irreversible tools (sign_transaction, send_transaction) lack documented confirmation mechanisms. No evidence of dry-run mode, confirmation step, or recovery guidance. Agents could sign/send unintended transactions without user approval.
Tool descriptions are minimal or missing. sign_transaction and send_transaction lack actionable descriptions explaining consequences (e.g., 'This is IRREVERSIBLE, funds will be sent to the blockchain'). No discovery of what these tools do beyond the name.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 35 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
No visible error handling or recovery guidance in source code. If call_contract_function fails due to invalid ABI or network error, no guidance on what to do next (e.g., 'Try get_contract_abi() first' for invalid ABI errors).
call_contract_function name is ambiguous. Does it call read-only views or state-changing functions? Agents may attempt to use it for writes when it only reads, or vice versa. Suggest 'call_contract_view' + 'call_contract_write' for clarity.
No documented output schemas. Agents cannot plan downstream calls or extract needed fields (e.g., transaction hash from send_transaction, balance from get_balance). Without documented return types, LLMs guess at field names.