Reports and Monitors

Payment wires: every paid endpoint accepts x402 and MPP (Machine Payments Protocol) on the same 402 - see Paying with x402 and Paying with MPP. Agent402 is the applied layer of Agentic Finance: agents that pay and get paid on their own.

Beside the 500+ per-call tools, Agent402 sells outcome-priced report products (one payment, one finished report) and monthly monitors that re-run a report when something changes. The same report endpoint serves two front doors:

• Agents call the /v1/... route directly and pay on the 402 - USDC over x402 or MPP.
• People buy the same report by card at /reports or subscribe to a monitor at /monitors.

A prepaid credits key also pays the agent routes (see "Credits keys" below).

All of it is optional and env-gated on a self-hosted instance (see Self-Hosting); without the Stripe keys the /v1 report routes still sell over x402 / MPP, and the card pages simply are not mounted.

Report products (agent path)

Each report is a POST with a JSON body; the price is the whole outcome, not a per-call meter. A report takes up to three minutes: send the paid call with Prefer: respond-async to get a job link at once and collect the report from it (see API-Reference, Async jobs), or send Prefer: wait=<maxSeconds> and keep the connection open that long. A paid call with neither header is answered 400 before any payment is taken. Either way the payment settles only on a delivered report. The evidence stage is deterministic (live web search, SEC EDGAR, openFDA, DNS/TLS probes, on-chain reads), and the synthesis stage writes a cited report that is checked against the evidence it was given; the IPO digest has no synthesis stage at all. Every model-written report's JSON also carries a grounding object: verdict (ok, unsupported_numbers or unchecked), how many figures were checked, and up to 25 figures the check could not trace to that evidence, each quoted in context. A figure the report derived (a sum or a difference) can be listed too, so read the list as figures to verify, not as errors. A run that cannot gather enough evidence fails with an error instead of shipping a thin report, and because settlement happens after the handler, an error is not charged on x402 (see Architecture).

Every report route is wallet-only (no proof-of-work tier) and is listed in /api/pricing, /openapi.json and on its own /tools/{slug} page with a sample output. They are not cached: each call is a fresh run.

# Unpaid, this answers the 402 quote; any x402 or MPP client pays it the same
# way it pays a $0.001 tool, then retries the same request.
curl -i -X POST https://agent402.tools/v1/domain-audit \
  -H 'Content-Type: application/json' \
  -d '{"domain":"example.com"}'

Over the MCP connector, catalog.find resolves a task like "audit example.com's email security" to the report slug and catalog.call runs it; the hosted connector can be paid over MPP right in the call (see MCP Connector).

Reports for people (/reports)

/reports sells the same products by card through Stripe Checkout: pick a product, type the input (a question, a ticker, a fund, a domain, a recall term, an article topic), pay, and the report renders at a private link /r/<session> (also emailed when the instance has email configured). The card price is derived from the agent price: $2 for the base tier (research, fund report, domain audit, token brief, recall report, insider report), $3 for the pro tier (research pro, market brief, dossier, fund report deep, domain audit pro, filing report), $4 for the max tier (research max, dossier max, LinkedIn article) and $5 for the ticker pack.

The card price includes payment processing. An agent paying per call over x402 or MPP pays the lower tool price in the table above for the same report. The token-risk and IPO reports are agent-facing routes and are not on the card page.

Guarantees enforced in code rather than promised in copy:

• No report is generated without a Stripe-verified paid session, and a session generates once (reloading the page re-serves the same report).
• A run that fails is refunded automatically; a refund that could not be issued is recorded as owed and retried, never reported as done.
• Report pages are private links (noindex, no listing anywhere); the link is the bearer.

Monitors (/monitors, $5 per month each)

A monitor is a Stripe subscription that keeps one report fresh for one target. The scheduler (src/monitor-scheduler.js) does a cheap, free check on a cadence and spends a full paid run only when the cadence says so or something actually changed, then emails a durable report link (/m/<id>).

Operational bounds: a target is validated at checkout (the domain must parse, the manager must resolve on EDGAR); paid re-runs are capped per subscription per 30 days (alerts keep coming); a failing run backs off (1h doubling to 24h) and never sends an email, and a target that keeps failing tells the subscriber once. Billing is Stripe's recurring invoice and the Customer Portal (the manage link in every monitor email) cancels or updates it; subscription status is re-read from Stripe before every paid run, so a canceled monitor stops being fulfilled.

Credits keys

Prepaid credits are a prepayment for Agent402's own services, for business and developer use; they are not transferable or redeemable for cash, never expire and carry no fees. /credits says whether packs are on sale (on a self-hosted instance, sales are switched on with CREDITS_SALES=on). A key (a402_…) pays per call:

curl -H 'Authorization: Bearer a402_…' 'https://agent402.tools/api/whois?domain=example.com'
# 200 … X-Credits-Balance: <what is left>   (list price held before the call, debited only on a final 200)

• Debited only on success. The gate holds the list price before the handler runs and converts the hold to a debit only when the final response is 200; any error or settlement failure releases it, and so does a client abort while the call holds a hang-up forgiveness ticket. Amounts are exact to the micro-dollar.
• Balance: GET /api/credits/balance (same header) returns what is left. An insufficient or unknown key answers 402 with { reason, balanceUsd }.
• Identity-bound tools refuse credits. The memory tools, my-usage, receipts, attest and feedback use the payer's signed wallet as identity; a credits key carries no verified wallet, so those routes answer 402 with reason: "identity-bound" and must be paid over an x402 rail.
• Outside-seller routes refuse credits. The route-execute tiers, seller-payability, /api/decide/execute and the flight routes buy from an outside seller with our wallet, so they answer 402 with reason: "third-party-purchase" and take a wallet payment only (x402, or MPP on a chain).
• In the SDKs: agent402-mcp (0.13.0+) reads AGENT402_CREDITS_KEY; agent402-client (0.7.0+) accepts { creditsKey }. Both keep their per-call and budget caps in front of the key.

How this fits the rest

Related: Tool Catalog · Paying with x402 · Paying with MPP · MCP Connector · Self-Hosting (the env that enables the card paths).