The FICO Score for AI agents.

AI agents can already research, negotiate, and pay. What they cannot do reliably is be trusted.

teranode is the credit bureau and underwriter that closes that gap. Every agent gets a score. Merchants gate on it. We monetize the graph, not the payments.

• The score. A signed 0–850 number + tier, built from transaction history, dispute rate, fraud rate, and principal lineage. One API call away from any merchant surface.
• The wedge. Agent payments over x402 on Base — how we earn the transaction data the score depends on. Live today, domain registration first.
• The moat. Once merchants require a teranode score to accept an agent, the score history itself is the asset. It compounds. It can't be re-created.
• The endgame. Underwrite fraud liability on scored agents. Lloyd's of London for the agentic economy.

Every merchant on the internet is about to be hit with agent-initiated purchases. There is no credit bureau to call.

• Merchants have no trust signal. Is this agent legitimate? Does it pay disputes? Who actually owns it? Today: no answer.
• Agent operators can't establish reputation. A well-behaved agent looks identical to a fraudulent one until it defrauds you.
• Underwriters can't price risk. No data, no premiums, no coverage — so every merchant eats agent fraud as pure loss.
• Payment rails are not the bottleneck. x402, USDC, and card APIs all work. The missing layer is the trust layer.

Humans solved this with FICO. Merchants can't accept plastic without pulling a score. Agents need the same primitive — and nobody is building it.

Six primitives — the infrastructure that generates the score data and exposes it to merchants.

Two common questions collapse into one answer once you see the layering.

• Stripe can build an x402 endpoint in a weekend. Stripe cannot build our score graph in five years.
• Every agent we process today becomes scored history. Every scored history makes the next merchant more likely to require a teranode score.
• The Visa analogy everyone uses is wrong. Visa is the rail. FICO is the chokepoint. We are building FICO.

Our first production merchant flow is domain registration. It is live today, and it's designed to seed the score graph.

• Why domain purchase. Universally understandable, trivial to demonstrate, real prerequisite for autonomous deployment. Every agent that ships something needs a domain.
• Why it matters for the score. Each registration exercises every primitive: x402 paywall, payment intent, spending mandate, signed attestation, ledger write. Every registration is a scored transaction.
• Merchant of record. teranode registers on behalf of the agent via name.com. We take a merchant fee; the agent never touches the name.com API. Revenue starts on transaction one.
• Status. Fully operational. Sandbox today; mainnet is one env var and one production API token swap away.

The wedge proves the primitives. The primitives are not domain-specific. Every subsequent merchant integration inherits the same substrate and feeds the same graph.

The credit-bureau model scales laterally. The same teranode score unlocks adjacent markets that a payments company could never touch.

• Agent KYC. Regulators are asking who the agent is. The score plus ERC-8004 lineage is the auditable answer.
• Agent employment verification. Businesses hiring agent workers need a reference check. That's a teranode score pull.
• Agent-to-agent contract underwriting. When agents contract each other, counterparty risk is priced off the score.
• Insurance pricing. Our own underwriting product prices premiums off the score. Third-party insurers license the score API for the same reason.
• Merchant enablement SDK. Any merchant gets a single integration: score pull, decision, settle. LangChain, CrewAI, AutoGPT framework plugins funnel every agent through the score.

The business is not a domain-purchase product. It is a credit bureau that starts at payments and expands to every market where agent trust is priced.

This is a working product, not a slide-deck concept. The infrastructure that generates the score data is live.

• x402 HTTP paywall — real on-chain USDC settlement via the Coinbase x402 facilitator, signed header validation, sub-second settlement on Base.
• ERC-8004 agent identity — on-chain principal resolution, registration params, reputation reader. Every transaction is attributed to a did:erc8004.
• SIWE wallet verification — EIP-4361 nonce generation, cryptographic signature verification, persistent sessions across all storage backends.
• Payment intent state machine — explicit lifecycle, transition guards, idempotency, reconciliation on any inconsistency. Every transition is signed evidence for the score.
• Spending mandates — per-call and cumulative ceilings, validity windows, persistent across sessions. Merchant-visible; scored inputs.
• name.com merchant-of-record flow — live reseller API, production credentials issued, sandbox default with a one-env-var flip to mainnet.
• Webhooks — HMAC-SHA256 signatures, 3-attempt retry with backoff, full delivery log per endpoint.
• Merchant API auth + admin — Bearer tokens, env-based and Postgres-backed credentials, timing-safe admin session.
• Multi-backend storage — InMemory / File / Postgres across every repository. Swap via one env var.
• Autonomous agent demo — Claude (claude-sonnet-4-6), MetaMask SIWE, SSE streaming, real x402 signing, persistent spending mandates.
• Discovery surfaces — A2A 0.3 agent card, Google AP2 manifest, MCP server with paid tools, llms.txt. Every agent runtime can self-configure.

Sandbox today. Mainnet is one env var: X402_NETWORK=base. Production name.com credentials are already issued.

The score-graph substrate is complete. What remains is turning the attestation layer into a live score and expanding into underwriting.