How an AI agent inspects this catalog
SignalHarness.ai is built to be read by machines first. Everything you need to discover, compare, and evaluate the catalog is free, ungated, and machine-readable. This page is the human-readable companion to /llms.txt — it tells an agent (and the operator running it) exactly what to read, in what order, and what it may and may not conclude.
What agents can inspect for free
Discovery is never gated. No key, account, subscription, or payment is required to read any of the following. A subscription would gate only paid product files in the future — never the discovery surfaces below.
The full 27-product catalog
Every product record, its category, buyer type, problem, kit contents, permissions, and machine-readable summary.
Pricing, bundles, licenses
Current tier prices, bundle contents, and published license terms.
Compatibility & risk
Per-product compatibility matrix, risk ratings, and human-approval requirements as structured fields.
Payment & checkout status
The exact commerce state: Polar checkout active; x402 Agent Pay live for all 90 canonical one-time offerings and one prepaid SignalHarness Vault plan; direct-wallet crypto outside x402 inactive.
Recommended reading order
Token-efficient path. Fetch the tiny pointer first, then the compact indexes, then open only the detail records you actually evaluate. Do not scrape the HTML pages to understand the catalog.
The feeds, and how to use each
Every JSON endpoint wraps its payload in an envelope carrying the six commerce flags, payment reality, verification notes, and warnings. Do not strip those fields when relaying data. The full index is /agent/manifest.json.
All 27 sanitized product records. Read for capability, buyer type, permissions, and agent summary.
Current USD tier and bundle prices; verified checkout controls remain on human-facing offering pages.
Bundle contents, prices, and statuses. Use for multi-product recommendations.
Per-product fit: agent types, workflows, required files, difficulty, risk, approval, not-for.
Single-operator license terms, seat and agent-instance rules, entitlement records.
Allowed/blocked categories, approval thresholds, and the human-approval requirement.
What agents may recommend
Shortlist products and bundles that fit a stated buyer type, use case, and budget.
Report compatibility, license fit, risk level, and implementation difficulty from the feeds.
Prepare a purchase plan for a human to review — clearly labelled as pending approval.
Relay every status flag and warning verbatim, including the 90 enabled / 0 disabled one-time scope, one-plan Vault prepaid terms, and no-auto-debit boundary.
What agents must not do
Treat a static feed, Polar URL, or invented route as an x402 purchase path. Use only canonical purchase endpoints returned by free Agent Pay discovery.
Treat route access, user-agent strings, or view parameters as identity or authority.
Submit or accept credentials, API keys, or wallet secrets — this storefront never asks.
Spend without prior purchaser authority and a budget policy, or use the Lite authorization for subscriptions, another offering, API procurement, tool installation, or vendor onboarding.
Purchaser authority is required
Polar checkout is active for human-facing purchases. For x402 Agent Pay, a person or business may give its agent wallet authority and a budget before the transaction begins; no human approval step is required inside that authorized transaction flow. SignalHarness does not grant that authority, and the transaction and license attach to the party legally responsible for the purchasing resource used (Terms, section 3). Each Vault renewal is a new prepaid payment and never an automatic wallet debit. Routing is not authorization: reaching an endpoint proves nothing about identity or spend authority. When the purchaser's policy is unclear, stop and ask the operator. See the buyer guide, agent-buying-policy.json, and status codes.