x402 agent-native · pay-per-call · signed receipts
A safety loop for every consequential agent action.
Call the one-cent gate before shell, Git, SQL, infrastructure, deploy, x402, or MCP actions. It returns deterministic risk and the exact next call; high and critical actions fail closed into a complete review-and-receipt workflow. Validation and finance utilities remain available behind the same contract.
New · flat monthly
Agent API Pass — $29/month, every endpoint
One subscription adds $35 of public-price-weighted credits across all 17 endpoints. A $0.01 gate consumes one cent and a $2 analysis consumes $2, so premium calls cannot arbitrage a raw-call cap. Per-call x402 continues after the credit pool is used.
Why agents put deterministic services inside the execution loop
A stable policy boundary across model updates
The model can propose an action; deterministic code decides whether it is low, medium, high, or critical. The same command produces the same matched rules regardless of which model proposes it.
No data feed to own
These are pure calculations over inputs your agent already has. Nothing to license, no rate limits, no keys — just a result and the evidence behind it.
Pay only when you call
x402 lets your agent settle a few cents per call autonomously over USDC on Base. No subscription, no onboarding — discover the endpoint and pay.
We benchmark this honestly: newer frontier models mostly pass our finance-math problems — while older production models returned a 279% IRR and missed a fund waterfall by $2,000,000. What no model gives you is the same answer after every version bump, or a signed record a third party can verify. That is what the deterministic engines and Assurance Attest are for.
Enrichment & validation — built for the loop
The agent-commerce pattern that works is one integration calling thousands of times. These per-record validators are built for exactly that: POST one record, get one deterministic verdict with evidence, repeat for the whole list. Validate and normalize a contact list before outreach; screen every counterparty wallet before your agent transacts.
EmailGuard
$0.02/callValidate & normalize every contact record your agent touches.
Loop-callable- RFC 5322, disposable, role & free-provider checks per record
- Typo fix (gnail.com → gmail.com) + canonical normalization
- Same record, same answer — loop whole lists at $0.02/record
POST /api/agent-services/emailguard/validateView examples & schema →WalletGuard
$0.10/callPer-wallet AML screening inside your agent's loop.
Loop-callable- Sanctions, exposure & fund-source signals
- 0–100 risk score with severity level
- Evidence-backed findings, never fabricated
POST /api/agent-services/walletguard/scoreView examples & schema →The trust layer for agent commerce
Decide whether to pay, inspect untrusted MCP tools, score destructive commands, route risky actions to review, and validate contact data before an autonomous agent acts. Tap any card for example requests and the schema.
Agent Action Safety Quick Gate
$0.01/callA one-cent safety gate for every consequential agent action.
Loop-callable- Deterministic low / medium / high / critical risk in one call
- Exact next-call routing with price, reason, and pre-mapped input
- Designed for high-frequency use before shell, Git, SQL, and deploy actions
POST /api/agent-services/action-safety/quickView examples & schema →Agent Action Safety Loop
$0.25/callScore, triage, receipt, and status in one agent-safety workflow.
Loop-callable- Full DestructGuard evidence with conditional ReviewQueue triage
- Honest allow / allow_with_controls / needs_review decision
- Signed receipt plus a free privacy-preserving 30-day status record
POST /api/agent-services/action-safety/runView examples & schema →x402 Demand Provenance Check
$0.05/callSeparate candidate demand from owner-funded and duplicate x402 activity.
Per-analysis- Excludes duplicate transaction hashes and declared owner wallets
- Measures payer independence, concentration, coverage, and freshness
- Caller-supplied evidence stays explicitly unverified
POST /api/agent-services/x402-provenance/checkView examples & schema →Assurance Attest
$0.05/callA signed, verifiable receipt for every action your agent takes.
Loop-callable- Binds an x402 payment or MCP tool call to a canonical HMAC-signed record
- Optional policy evaluation and settlement outcome in the same record
- Free public verify endpoint — third parties check it without trusting your logs
POST /api/agent-services/assurance/attestView examples & schema →Payment Policy Decision API
$0.05/callDecide whether an agent should pay before it signs.
Loop-callable- Checks URL, amount, network, payTo, facilitator, and buyer limits
- Flags resource-binding mismatches, replay gaps, PII leaks, and price anomalies
- Returns allow / deny / needs_review with redacted log-safe evidence
POST /api/agent-services/payment-policy/decideView examples & schema →Evidence Verification API
$0.05/callVerify the evidence bundle before trusting a paid agent result.
Loop-callable- Checks x402 challenge, policy decision, receipt, and service response artifacts
- Rejects endpoint, method, network, payTo, and price mismatches
- Returns a run-log policy record with per-rule evidence
POST /api/agent-services/evidence/verifyView examples & schema →MCP Server Risk Analyzer
$0.05/callScore an MCP server before your agent installs it.
Loop-callable- Destructive tools, over-broad scopes, weak auth
- Per-tool, per-factor findings
- A concrete fix, not just a number
POST /api/agent-services/mcp-risk/scoreView examples & schema →DestructGuard Command Score
$0.10/callCatch destructive agent commands before they run.
Loop-callable- Scores shell / SQL / git actions for risk
- Flags force-push, hard-reset, drops, teardown
- Returns an approve-or-block recommendation
POST /api/agent-services/destructguard/scoreView examples & schema →ReviewQueue Agent Submit
$0.25/callRoute risky agent actions to human approval.
Loop-callable- Paid risk triage + approval routing
- Request id and recommended action
- Pause-and-escalate on critical risk
POST /api/agent-services/reviewqueue/submitView examples & schema →How an agent calls a service
- 1. POST your inputs. The first unpaid request returns HTTP 402 with an x402 payment challenge.
- 2. Pay per call. Your x402 client signs a USDC-on-Base payment and retries — no API key, no account.
- 3. Get a result + evidence. HTTP 200 returns the computed values plus per-rule findings explaining them.
- Endpoints are published in the OpenAPI spec and x402 manifest; unpaid 402 challenges include Bazaar metadata so agents can inspect inputs, outputs, and price before signing.
POST https://hermesplant.com/api/agent-services/cashflowlens/analyze
{
"cashflows": [-1000000, 50000, 60000, 70000, 1200000],
"discountRate": 0.08
}
→ 402 Payment Required (x402 challenge)
→ sign USDC payment, retry
→ 200 OK
{
"summary": { "irrAnnual": 0.0966, "npv": 110446.12, "moic": 1.38 },
"findings": [ ... ]
}Integrate in minutes
Start free with an API key (no wallet), or pay per call over x402 — zero-code over MCP, the x402 fetch client, or raw HTTP.
Start free — no wallet (60 seconds)
Get a free key (250 calls/mo, every endpoint) and send it as X-API-Key. Built to run in your agent’s loop — one call per record, the same audited answer every time.
# get a key (returned once), then loop over your records
curl -s -X POST https://hermesplant.com/api/keys/free \
-H "content-type: application/json" -d '{"email":"you@company.com"}'
curl -s https://hermesplant.com/api/agent-services/emailguard/validate \
-H "X-API-Key: hp_free_..." -H "content-type: application/json" \
-d '{"email":"alice@gnail.com"}'Full quickstart in the API docs or the npm SDK. Past 250/mo, upgrade to the $29 Pass or keep paying per call over x402.
1. Zero-code via MCP
Add the Hermes Plant MCP server to any MCP client (Claude Desktop, Cursor, your agent framework). The finance tools appear automatically; calls return a 402 your client settles with an x402 wallet.
# MCP server (streamable HTTP)
https://hermesplant.com/mcp
# tools exposed: cashflowlens_analyze, waterfall_distribute,
# options_price, bond_analyze, portfolioguard_score, walletguard_score, ...The server is published on Smithery, mcp.so, and Glama, so MCP clients can inspect the exact published listing before connecting.
2. The x402 fetch client (Node / TypeScript)
@x402/fetch wrapsfetch so the 402 → pay → retry handshake is automatic. Your agent just calls the endpoint.
import { wrapFetchWithPayment } from "@x402/fetch";
import { privateKeyToAccount } from "viem/accounts";
const account = privateKeyToAccount(process.env.AGENT_WALLET_KEY);
const pay = wrapFetchWithPayment(fetch, account); // USDC on Base
const res = await pay("https://hermesplant.com/api/agent-services/cashflowlens/analyze", {
method: "POST",
headers: { "content-type": "application/json" },
body: JSON.stringify({
cashflows: [-1000000, 50000, 60000, 70000, 1200000],
discountRate: 0.08,
}),
});
const { summary } = await res.json();
console.log(summary.irrAnnual, summary.npv, summary.moic);3. Inspect the 402 with curl
An unpaid request returns the payment challenge — price, asset, network, and recipient.
curl -i -X POST https://hermesplant.com/api/agent-services/options/price \
-H "content-type: application/json" \
-d '{"spot":100,"strike":100,"timeToExpiry":1,"volatility":0.2,"riskFreeRate":0.05}'
# → HTTP/1.1 402 Payment Required (x402 challenge in the payment-required header)Finance and deterministic math
The same pay-per-call model for calculations agents should not improvise: returns, waterfalls, derivatives, bonds, portfolio risk, cashflow analysis, and underwriting.
CashflowLens
$0.20/callNPV, IRR, XIRR & DCF valuation in a single call.
Per-analysis- Solve IRR / XIRR from any cashflow series
- NPV + DCF (Gordon-growth or exit-multiple)
- MOIC / DPI / TVPI and payback period
POST /api/agent-services/cashflowlens/analyzeView examples & schema →WaterfallLens
$0.25/callLP/GP distribution waterfalls, solved exactly.
Per-analysis- Return of capital → preferred → catch-up → carry
- Exact effective carry %, per-tier breakdown
- Compounded or simple preferred return
POST /api/agent-services/waterfall/distributeView examples & schema →OptionLens
$0.25/callBlack-Scholes option pricing with the full Greeks.
Per-analysis- Calls & puts with continuous dividend yield
- Delta, gamma, vega, theta, rho
- Intrinsic / time value and moneyness
POST /api/agent-services/options/priceView examples & schema →BondLens
$0.25/callYield, duration, convexity & loan amortization.
Per-analysis- Price ↔ yield-to-maturity
- Macaulay & modified duration, convexity
- Level-payment loan amortization schedules
POST /api/agent-services/bond/analyzeView examples & schema →PortfolioGuard
$0.15/callPortfolio risk scored straight from holdings.
Per-analysis- Volatility, Sharpe, max drawdown
- Concentration (HHI) & diversification
- Per-rule findings with concrete fixes
POST /api/agent-services/portfolioguard/scoreView examples & schema →DealAnalyzer
$0.25/callFull deal underwriting — DCF, returns, sensitivity & waterfall in one call.
Per-analysis- DCF valuation: enterprise value, equity, share price
- IRR/XIRR, NPV, MOIC + a sensitivity grid
- LP/GP waterfall on the DCF proceeds
POST /api/agent-services/dealanalyzer/analyzeView examples & schema →Frequently asked questions
- How do AI agents pay for Hermes Plant agent services?
- Each endpoint is metered with the x402 protocol. An agent's first request returns an HTTP 402 challenge; the agent signs a small USDC-on-Base payment and retries to get the result. There is no API key, account, or subscription — the agent pays a few cents per call autonomously.
- Why call an API instead of letting the LLM compute the math?
- Two reasons that survive model progress: stability and proof. Even when a frontier model computes correctly, its answers can change across versions and runs — our benchmark measured run-to-run instability even in models that scored 8/8 on accuracy. Hermes Plant endpoints are pinned, deterministic, unit-tested functions: the same inputs always return the same audited result, with per-rule findings, and any action can be sealed with an HMAC-signed attestation a third party can verify.
- Can my agent prove what it did and what it paid?
- Yes. Assurance Attest ($0.05 per call) binds an x402 payment or MCP tool call — including its settlement outcome — into a canonical, HMAC-signed record. Anyone can confirm the record at the free /api/agent-services/assurance/verify endpoint without trusting the agent's own logs. Assurance Team and Business plans add organization-wide decision history, receipts, and drift probes.
- Do these finance APIs require a market-data feed or API key?
- No. They are pure calculations over the inputs your agent already has — cashflows, holdings, option parameters, bond terms. There is nothing to license, no rate limits, and no keys to manage.
- What agent trust and finance services are available?
- Signed action attestations, payment policy decisions, MCP server risk analysis, destructive-command scoring, approval routing, email validation, DCF and IRR/XIRR, LP/GP waterfalls, Black-Scholes option pricing, bond analytics, portfolio risk scoring, and wallet evidence scoring are available as x402-paid agent services.
- How much does each call cost?
- Per-call prices range from $0.01 to $0.25 in USDC on Base. You pay only when your agent calls the endpoint — no minimums and no subscription.
- How does an AI agent discover these endpoints?
- They are published in the OpenAPI spec and /.well-known/x402 manifest. Each unpaid 402 challenge carries Bazaar discovery metadata for the endpoint, so agents can inspect the inputs, outputs, price, and payment terms before signing.
- Are the results auditable?
- Yes, twice over. Every response includes evidence-backed findings — which rules fired, why, and a concrete fix or note. And every paid response carries a signed receipt: the endpoint plus a SHA-256 hash of the exact response body, HMAC-signed, verifiable for free at /api/agent-services/assurance/verify. Nothing is fabricated.
- Can I request a calculator that isn't in the suite yet?
- Yes. Contact us and we will add deterministic endpoints to the finance and quant suite.
Wire a service into your agent today
Start from API docs and endpoint schemas, then choose the right commercial model for your traffic. If you need a custom calculator, we can scope that too.
Questions? contact@hermesplant.com