Loading ToolBench
MCP client for NanoParse — pay-per-parse web content extraction for AI agents
Single tool with clear naming (nanoparse_fetch) and well-structured input schema. Description is adequate (194 chars) and explains the payment model. However, output schema is not formally documented, the response structure is inferred from code (markdown, metadata, debug fields) but not declared in the tool definition. Parameters are typed and described, but the tool lacks error handling guidance for payment failures, network errors, and API-specific edge cases. No tool annotations (readOnlyHint, idempotentHint) despite being a read-only operation.
Fetch a web page and return clean, structured Markdown with metadata. First 10 parses free per device. After that, pay $0.01 USDC on Base via x402. No account or API key required.
Output schema not formally documented. Response structure (markdown, metadata, debug, structuredContent) is inferred from code but not declared in tool definition. LLMs cannot plan downstream operations without knowing return fields.
No error handling guidance. Tool can fail with 402 payment required, network errors, invalid URLs, and API errors, but descriptions do not explain recovery steps or error categories (retryable vs fatal).
Missing tool annotations. Tool is read-only and idempotent but lacks readOnlyHint and idempotentHint annotations. LLMs cannot infer safety properties without explicit hints.
Payment flow complexity not surfaced in description. x402 payment requirement, free call limits, and wallet key setup are mentioned but not clearly explained as prerequisites or failure modes.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 69 | 2025-06-18+ | v2 |
No pagination or result limits documented. Tool returns full page markdown without documented limits. Large pages could exhaust context windows.