MCP server that exposes Moralis Web3 API endpoints as tools, converting OpenAPI specifications to MCP-compatible tools for blockchain data queries
This server dynamically generates tools from the Moralis OpenAPI spec. While the architecture is sound (MCPSchemaConverter, ToolRegistry, ParameterFilter classes), the actual tool definitions are NOT visible in the source code provided. Tool names, descriptions, and schemas are inferred from OpenAPI operationIds and parameters at runtime. Per HARD SCORING RULES, when tool definitions are inferred rather than directly visible, per-tool scores cap at 50. Additionally, the code shows Zod schema construction but does NOT expose the final inputSchema.properties structure, critical for validation. No output schemas are documented. The single tool entry is a placeholder description, not a concrete definition. This is a code-generation pattern, not a hand-crafted tool suite, which limits our ability to assess naming clarity, description quality, and parameter documentation against the 54 patterns.
Tools are dynamically generated from the Moralis OpenAPI specification. The server converts all non-blacklisted OpenAPI operations into MCP tools. Specific tool names and schemas are derived from the operationId and parameters defined in the Moralis API specification (moralis.ts). Examples include blockchain data queries, token analytics, NFT metadata, transaction history, and wallet information endpoints.
Tool definitions are dynamically generated and not visible in source code. Cannot verify naming follows verb_noun convention, descriptions are LLM-optimized, or schemas are complete. Per HARD SCORING RULES, inferred tools cap at 50 overall.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields. Each dynamically generated tool lacks documented return types.
Parameter descriptions are sourced from OpenAPI spec but not validated for LLM clarity. OpenAPI descriptions often lack context on WHEN to use a tool, prerequisites, or state-modification warnings required by pattern:command-tool.
No error handling guidance visible. Tools return raw API responses without recovery hints, categorization (retryable/user-fixable/fatal), or actionable error messages. Agents cannot self-correct on failures.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 29 | 2026-07-28+ | v2 |
No pagination or result-limiting logic visible. Moralis API may return large datasets; without cap and pagination, responses could exhaust context windows and degrade LLM reasoning.