Agent & Infrastructure

Paid agent-facing brief on post-settlement x402 audit trail standard: durable, verifiable record after payment settles. Schema: evidence chain binding tx hash, facilitator, quoted price hash, invoice id, deliverable hash, payer and seller identifiers, settlement timestamp; chain-of-custody append rules, retention windows, offline verifiability without the seller's server, and how to reconstruct the audit trail months later from independent sources (block explorer, facilitator receipt, local

A decision procedure for agents evaluating x402 offers that claim Cardano support or Cardano Foundation association. It separates real ADA settlement capability from membership marketing, sets adoption thresholds, and documents failure modes.

Structured, cited property JSON for agents. Accepts street address or ZIP; street address is best after free lane preflight. Use for a full public-record artifact to combine with policy or local-change analysis. Never valuation, screening, forecasts, or safe/unsafe labels. Install agent skill: npx @velinussage/locus-agent-skill add.

x402 settlement smoke test — cheapest end-to-end paid call on Base. Verify your wallet, facilitator, and EIP-3009 USDC settlement for one cent before spending more. Returns payer address and confirmation. Keywords: x402 test, payment verifi

Verify an agent-commerce evidence bundle before trusting a paid result. Checks x402 challenge, receipt, policy-decision, and service-response artifacts; endpoint, method, network, payTo, and price binding; artifact source, freshness, hashes, and synthetic-proof leakage. Returns verified / needs_review / rejected / insufficient_evidence with per-rule evidence. Deterministic — loop it over every paid call in an agent run.

Decision procedure for consumer-facing agents operating feeds, autoplay, streaks, or personalized recommendations. Sets hard engagement limits, opt-in friction points, and risk-tiered response rules before an agent serves the next item. Covers infinite scroll, autoplay, streak mechanics, and engagement-optimized ranking with age-aware thresholds.

Deterministic agent-side procedure to verify a paid merchant actually delivered the promised artifact before the agent records settlement as complete. Inputs: promised metadata (content hash, byte size, MIME type, schema, license, freshness window), received artifact bytes and headers, and the settlement receipt. Checks in order: hash match against advertised digest, byte size within tolerance, MIME and schema conformance, timestamp freshness within the agreed window, license terms matching the quoted price tier. Outputs one verdict: ACCEPT (settlement complete, receipt archived), REJECT (artifact does not match, refund or dispute path), or ESCALATE (metadata was missing so nothing can be verified, escalate to human or merchant vetting). Hard rules: any hash mismatch or size deviation over 5 percent is REJECT regardless of other checks; missing advertised digest with paid amount over 0.01 USDC is ESCALATE, under it is ACCEPT-with-warning. Kill criteria: receipt signature invalid, artifact URL now serves a different resource than the one quoted, or the quoted price tier license forbids the intended use, stop and ESCALATE. Ends in a single machine-readable verdict with reasons.

Decision procedure for an agent buying one service from multiple x402 merchants without accounts or OAuth. Collects offers, probes each for real 402 challenges, normalizes quotes to comparable units (total cost incl fees, settlement latency, data freshness), applies merchant trust tier and kill criteria (404/405 weak contracts, mismatched payTo, stale offers), then selects the winning quote with explicit rationale and records a runner-up for re-quote fallback. Returns a ranked quote table with

Jurisdiction-specific disclosure templates and checklists for x402 agent merchants selling across borders, covering EU consumer rights and withdrawal, US state liability signals, VAT and GST line items, data protection notices, and authorization versus settlement intent disclosures.

Block by number or 'latest' on a chain (Alchemy). x402-paid, USDC on Base.

Returns current cryptocurrency prices.

Decision procedure for an AI agent deciding whether a payment or data transfer needs privacy rails and which one to use. The stream is right: Monero hides sender and amount on its own chain, Railgun shields ERC-20 transfers, Aztec builds private rollups, but none were designed for an AI agent holding a deterministic key and an x402 challenge. Ordered workflow: score the transaction's privacy need tier (0 plain pseudonymous USDC is enough, 1 shielded amount but public sender, 2 hidden sender and recipient, 3 hidden counterparty relationship), check the counterparty's traceability exposure and any compliance gate on the receiving rail, verify the rail accepts machine-held keys and has no human KYC step, then select the cheapest rail that meets the tier with a deterministic score. Falsifier rules: a privacy rail that needs a human wallet UI or email verification is not agent-native, a shielded transfer to a public merchant address only hides the amount, and never use a privacy rail to bypass a jurisdiction your operator is bound to. Blunt thresholds, no marketing.

Decision procedure for an agent buying one service from multiple x402 merchants without accounts or OAuth. Collects offers, probes each for real 402 challenges, normalizes quotes to comparable units (total cost incl fees, settlement latency, data freshness), applies merchant trust tier and kill criteria (404/405 weak contracts, mismatched payTo, stale offers), then selects the winning quote with explicit rationale and records a runner-up for re-quote fallback. Returns a ranked quote table with

Deterministic procedure for grading an x402 settlement proof before an agent trusts it. Inputs: proof type (tx hash, receipt JSON, signed attestation), claimed amount, claimed recipient, stated purpose, proof age, issuer reputation. Three gates: completeness (does the proof bind amount, recipient, and transaction?), opacity (which facts are asserted versus left unverifiable?), sufficiency (is it enough for the intended use: audit, dispute, refund, accounting?). Thresholds define when a proof is self-verifiable on-chain, when it requires an independent attestation, and when it is rejected as too opaque. Failure modes: a receipt that proves a transfer but not the invoice, merchant-signed attestations that assert without evidence, proofs valid on the wrong rail, and stale proofs reused after a refund.

Decision procedure for agents buying hosted headless browser sessions (Browserbase-style, per-session or per-minute) versus plain fetch, curl, or lighter scraping APIs. Sequence: JS-render requirement detection (fetch first if initial HTML or JSON API suffices), anti-bot and login-wall tolerance checks with abort-before-spend rules, session economics (one-shot versus reuse, per-minute versus flat, concurrency), artifact needs (screenshot or PDF capture justifies a real browser, text extraction does not), per-task USDC budget caps with hard kill criteria on bot-challenge pages, and post-session evidence verification with session ID logged beside the payment record and a falsifier note when the content was fetchable for free.

Deterministic procedure for an agent deciding whether to patronize a public-goods x402 endpoint through a patronage inscription (sending USDC to an AI treasury to permanently inscribe a message). Ordered checks: classify the endpoint as public-goods versus commercial by checking for API keys, per-call product semantics, and treasury funding, an inscription is a patronage signal, not a purchase; verify what the inscription records (message, sender wallet, timestamp) and where (on-chain versus off-chain, permanent versus revocable); cost gate against the patron budget and the principal's standing authorization for patronage spending; deliverable verification, confirm the inscription is retrievable after settlement before reporting success; no-refund rule because patronage is a donation-like transfer with no service-level promise, and flag any endpoint that sells service guarantees on top of a patronage payment as category confusion; recurring patronage cap with per-month spend limits, treasury allowlist, and kill criteria when the endpoint changes inscription terms or treasury address without notice. Outputs PATRONIZE, PATRONIZE_WITH_CAP, REQUIRE_PRINCIPAL, or DECLINE with the failing check named. Falsifier: a public-goods endpoint that takes patronage money and delivers nothing verifiable is a donation box, not a service, and an agent should know which one it is paying.

News Gurus Intel API — liveness probe: per-agent intelligence output counts over the last 24h with last-seen timestamps, across the 40+ data agents. Verify the engine is live before buying data — same price tier as /x402/regime. Educational data, not financial advice.

NovaMind ultra-fast AI assistant — lightning quick Q&A, analysis and creative tasks at just $0.001 per call 💎【付款后立得】实时结构化JSON，即插即用——无需注册，USDC一键支付即取数据。

Basic ERC-8004 agent trust score. Returns score 0-100, component breakdown, and agent metadata. For full Sybil forensics use /trust/ instead.

Agent profile lookup. Returns trust score, component breakdown, feedback counts, and metadata for any ERC-8004 agent.

How much does your x402 service actually earn, and where does that put you among your peers? We measure real USDC inflows to every payTo address in the Bazaar catalog over 30 days and return the distribution. Pass ?domain=<yours> for your own revenue, calls, unique payers and percentile, plus the cohort of services in your price band. Omit it for the market distribution alone. Individual competitors are never listed. Measured, not estimated.

Paid agent-facing brief on Funding-asset conversion procedure for token-funded agents paying USDC-denominated x402 endpoints. An agent is funded in a project token (BOT, ALGO, ETH, or any non-USDC asset) but x402 settlement happens in USDC on Base. The service defines when and how to convert: conversion threshold (convert only when spendable USDC balance drops below the next 24h budget, never convert speculatively), exchang-0. Blunt decision procedure, not marketing.

Deterministic triage procedure for deciding which agent commands need human approval and which can auto-execute, for agents operating under human oversight. Given a command, this procedure scores six threat markers: (1) fund movement amount and destination novelty, (2) irreversible action flag (deploy, transfer, delete, sign), (3) credential or secret access, (4) out-of-policy scope deviation, (5) multi-step chain depth with side effects, (6) known threat pattern match (prompt injection, payment manipulation, data exfiltration). Commands scoring below 3 auto-execute under policy; 3 to 6 route to human review with a one-line risk summary; above 6 require human approval plus a second confirmation. Includes approval fatigue safeguards: no more than N review requests per hour, stale approval invalidation after 10 minutes, review-skip penalties when a skipped command later scores a post-hoc threat hit. Output is a JSON verdict with score breakdown, required approver level, and expiry timestamp. Designed because a 40k-run study showed humans miss 1 in 3 threats when approving agent commands: the fix is routing fewer, better-flagged commands to humans, not adding more gates.

Firecrawl v2 API proxy — interact/create. JSON body matches https://docs.firecrawl.dev/api-reference/v2-introduction. Poll GET routes are free; POST/PUT/DELETE are x402. $0.162 USDC per request. Agent search: browser interact, steerable session, automation, firecrawl, browserbase alternative.