Loading ToolBench
MCP tools for x402 preflight and claim verification through InferenceView
This server defines two tools with explicit schemas and descriptions, but both tools exhibit significant quality gaps. The tool names are action verbs (verify_transaction, factcheck), but descriptions lack detail on state changes, error handling, and when to use each tool. Parameter descriptions are minimal or absent. Output schemas are not documented, callers cannot predict response structure. The factcheck tool's 'depth' parameter has no description despite being an enum. Context object properties are documented as 'optional' without explaining what each field controls. No error guidance is provided. The server targets an external API (InferenceView) but provides no documentation on failure modes, retries, or cost implications of the x402 payment system. Overall, definitions are present but lack the LLM-optimized rigor required for agent reasoning.
Verify a factual claim before acting on it. Returns a verdict, confidence, primary sources, freshness, contradictions, and evidence.
Before paying an unknown x402/MCP resource, return risk, recommendation, and evidence. The live x402 challenge controls the price.
No output schema documentation. Both tools return JSON from external API without documenting field names, types, or structure. Agents cannot predict response format or plan downstream calls.
factcheck 'depth' enum parameter lacks description. Despite having an enum constraint (quick|standard|deep), there is no explanation of what each depth level does, when to use each, or what cost/latency differences exist.
factcheck 'context' object properties are undescribed. The url, company, person, product, data_point fields are marked optional but lack explanations of how each refines the factcheck or when to provide them.
No error recovery guidance. Both tools call external APIs (x402 payment, InferenceView factcheck) without documenting failure modes (network timeout, payment rejection, rate limits, API errors). Agents have no guidance on retry logic or fallback strategies.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
verify_transaction description does not explain state changes. The phrase 'return risk, recommendation, and evidence' hints at the output, but does not explain what 'risk' and 'recommendation' mean, what evidence format is returned, or whether the tool actually initiates a payment or only checks.
Missing cost/side-effect disclosure. The verify_transaction description mentions 'The live x402 challenge controls the price' but does not explicitly state that calling this tool triggers a payment or preflight check. Agents need to know if this call costs money.
No documented return structure or field mapping. External API returns JSON, but the tool definitions do not describe what fields callers should expect, what types they are, or how to extract actionable information for downstream decisions.
factcheck 'claim' parameter minLength=8 is undocumented. The constraint exists in schema but is not explained in the description, leaving agents uncertain why very short claims are rejected.