Loading ToolBench
Install one middleware. Monetize unused 402 responses with recommendations, discovery, and machine commerce analytics. MCP server: 7 tools for agent demand analytics, paid per call in USDC on Base; free for publishers on their own traffic.
Server defines 7 tools with explicit schemas via Zod and tool annotations present. Naming follows verb_noun pattern (get_*, preview_*). Descriptions are present and substantive (100-200 chars), explaining pricing and use cases. However, several tools have optional parameters without clear defaults, and output schemas are not documented, LLMs cannot predict response structure. Error handling for paid tools mentions x402 payment challenges but lacks actionable recovery guidance. Tool composition is sound (each does one thing), but parameter descriptions could be more prescriptive about constraints.
Paid, $0.02 USDC on Base via x402. Demand depth for one category: probe volume, distinct sources, buyer-class share, price points probed, daily series. Without WALLET_PRIVATE_KEY, returns the x402 payment challenge instead of settling.
Why-agents-didn't-buy funnel for one service: bounce funnel, traffic classes, top abandoned endpoints, retry signals. FREE with X402_ADS_PUBLISHER_KEY for your registered services (the service id is bound to your key at registration); otherwise $0.05 USDC on Base via x402. Without a publisher key or WALLET_PRIVATE_KEY, returns the x402 payment challenge.
Paid, $0.01 USDC on Base via x402. Google-Trends-for-agents: top requested x402 endpoints and categories by autonomous agents, split by traffic class. Without WALLET_PRIVATE_KEY, returns the x402 payment challenge instead of settling.
Free. Live totals for the ForgeMesh machine-commerce network: 402 responses observed, agent-class requests, recommendations served, services reporting, and x402 services indexed.
Paid, $0.005 USDC on Base via x402. Network-wide intent stats: total events, services, monitor/indexer/agent traffic classification split, and ad activity. Without WALLET_PRIVATE_KEY, returns the x402 payment challenge instead of settling.
Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination). Descriptions mention what data is returned but not the JSON shape.
Error handling for paid tools (get_network_stats, get_intent_trends, get_category_demand, get_intent_report) mentions x402 payment challenges but provides no actionable recovery guidance. LLM cannot determine if error is retryable, user-fixable, or fatal.
preview_recommendations has three optional parameters (service, endpoint, category) with no documented defaults or constraints. Unclear if all three are required together, mutually exclusive, or independent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | 2026-07-28+ | v2 |
Free. The network's canonical terms of service and complete data-collection disclosure: exactly what the middleware sends and never sends.
Free. See the exact typed recommendations block (sponsored + similar x402 services) that the @forgemeshlabs/x402-ads middleware would inject into a 402 response for a given endpoint and category.
Parameter descriptions lack prescriptive constraints. E.g., 'window' enum is documented but no default stated; 'limit' has min/max but no guidance on typical values or performance implications.
WALLET_PRIVATE_KEY and X402_ADS_PUBLISHER_KEY are environment variables, not tool parameters, correct for secret injection. However, no tool description explains how to configure these or what happens if missing (payment challenge vs. free tier).