Loading ToolBench
MCP server for the DYOE Agent Tools — pay-per-call trust, due-diligence, commerce & data tools for AI agents, settled in USDC on Base via x402. Includes the human-verified Guardian approval layer.
16 tools with clear, marketing-focused descriptions (avg 180 chars) and basic Zod schemas. However, critical gaps: (1) Most parameter descriptions are minimal (avg 30 chars, well below 72-char baseline). (2) No output schemas documented, LLMs cannot plan downstream calls or extract structured data. (3) No error handling guidance, tools return raw JSON with no recovery hints. (4) Parameter constraints missing, e.g., 'symbol' accepts any string, no enum of valid coins. (5) No idempotency or side-effect declarations despite write operations (guardian_approve). (6) Tool names lack action verbs (trust_check, merchant_trust, token_intel are noun-heavy; should be verify_trust, check_merchant, get_token_intel). Schemas are present but minimal; descriptions are marketing copy, not LLM-optimized.
Global crypto market overview — total market cap, 24h change, BTC/ETH dominance. Costs $0.01 USDC on Base.
Top gainers & losers (24h) across the top 100 coins — for trading agents. Costs $0.01 USDC on Base.
Live crypto price + 24h change, market cap & volume for any coin — for trading agents. Costs $0.01 USDC on Base.
The coins trending right now — for trading & research agents. Costs $0.01 USDC on Base.
Extract emails, phone numbers & social links from any website — lead enrichment & prospecting for agents. Costs $0.01 USDC on Base.
Extract clean main text content from any web page (no nav/ads) — for research & summarizing agents. Costs $0.01 USDC on Base.
No output schemas documented. LLMs cannot infer what fields to expect (e.g., trust_check returns 'verdict', 'trust_score', 'domain_age', 'red_flags', none declared). Breaks downstream tool chaining and forces agents to guess field names.
Parameter descriptions are minimal (avg 30 chars vs 72-char baseline). E.g., 'symbol' → 'Coin symbol, e.g. btc' lacks format, range, or valid values. No enums for constrained inputs (symbol, currency codes). Invites hallucinated values.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Currency conversion + live exchange rate (ECB data). Costs $0.01 USDC on Base.
HUMAN-VERIFIED approval for an agent action. Before an agent pays, signs, or acts for its human, a REAL PERSON signs off when it matters (plus automated trust + policy). Returns approved / pending_human / denied. The human-in-the-loop accountability layer agents can't get anywhere else. Costs $3.00 USDC on Base.
Full due-diligence dossier on any company or website in ONE call: what they do, legitimacy + trust verdict, domain age, contacts, social presence, scam red flags. Counterparty research & background checks for agents — one call instead of ten. Costs $0.10 USDC on Base.
Is this online store safe to buy from? Scam/fraud check + trust verdict (safe/caution/avoid) + store signals (checkout, returns, payment methods) — counterparty trust for shopping & commerce agents before they pay. Costs $0.05 USDC on Base.
Extract title, description, OpenGraph image, favicon & canonical from any URL — link previews & summaries. Costs $0.01 USDC on Base.
Is this website legit and trustworthy? Full trust + SEO + contactability audit scored 0-100 with a legitimacy verdict and prioritized fixes. Due diligence in one call. Costs $0.05 USDC on Base.
TRUST-VETTED token intelligence — price + market data PLUS a scam/rug-risk read on the project behind the coin (legit site? brand-new domain? red flags?). Only DYOE fuses market data with trust. Costs $0.05 USDC on Base.
Is this counterparty safe? Scam, fraud, phishing & rug-risk check before an agent trusts or pays any website, business, wallet, or endpoint. Returns verdict (safe/caution/avoid), trust_score, domain age & red_flags. The trust tollbooth of the agent economy. Costs $0.01 USDC on Base.
Current conditions + 3-day forecast for any city — for travel & planning agents. Costs $0.01 USDC on Base.
Factual summary of any topic from Wikipedia — for research agents. Costs $0.01 USDC on Base.
Tool names lack action verbs. 'trust_check', 'merchant_trust', 'token_intel' are noun-heavy; should be 'verify_trust', 'check_merchant_safety', 'get_token_intel'. LLMs infer intent from verb-first names.
No error handling guidance. Tools return raw JSON or error messages with no recovery hints. E.g., if 'symbol' is invalid, LLM gets no hint to call crypto_trending first. No categorization of retryable vs fatal errors.
guardian_approve is a write operation (human approval, payment) but has no idempotency declaration, confirmation step, or dry-run mode. Agents could retry and trigger duplicate approvals.