~/withgordon.ai/ai.md—● buffer modifiedutf-8 · lf · markdown
$ curl -s https://withgordon.ai/ai
● machine view · agent-readable mirror · updated 2026-06-18

# Gordon AI

> Gordon is the enterprise control plane for agentic commerce. It gives AI agents a compliant wallet, evaluates every payment intent against policy and a live risk model, signs the decision cryptographically, and reconciles the resulting invoice into your ERP.
[Book an intro]

## Overview

Gordon sits upstream of every payment rail. We intercept the agent's intent, evaluate it against policy and a live risk model, and sign a decision that travels with the transaction. The same control plane covers both agent-to-merchant (A2M) and agent-to-agent (A2A) flows.

## Use cases

### A2M · Agent → Merchant

An AI agent pays a human-owned merchant (e.g. chatgpt.travel → expedia.com). Gordon sits between the two, enforces the buyer's policy pre-execution, and emits a signed decision that travels with the transaction.

### A2A · Agent → Agent

Two AI agents transact (e.g. marketing.agent ↔ underwriting.agent). Gordon attributes the spend on both sides, audits the interaction, and routes settlement through x402 or a programmable rail.

## Three primitives

### 001Account — FDIC-insured agent wallets

Onboard agents with a wallet that can hold, receive, and send funds. Spend rules are policy-as-code, applied before the agent's tool call is allowed to execute.

  • - FDIC-insured wallets bound to a cryptographic agent identity
  • - Vendor allowlists, spend caps, and velocity controls
  • - Settlement via x402, Machine Payments Protocol (MPP), merchant tokens

### 002Fraud Detection & Audit Trails

Detect agent-native fraud at the decision layer (where the agent reasons), not at the checkout (where the dollar has already moved). Every decision is signed and logged immutably for compliance, finance, and security.

  • - Prompt injection and catalog poisoning detection
  • - Intent deviation scoring per transaction
  • - Cryptographically signed, append-only audit trail

### 003Invoice & Outcome Reconciliation

Cryptographic agent identities map every transaction to a scope, a permission, and the authenticated human behind it. Invoices auto-reconcile and sync directly into your ERP — Gordon becomes the system of record for agent spend.

  • - Cryptographic agent IDs with scoped permissions
  • - Auto-reconciliation of invoices to source intent
  • - Native sync into NetSuite and other ERPs

## Decision pipeline

Every transaction passes through four stages. Each stage is traceable in the audit trail; the cryptographic signature produced in step 3 is attached to the on-chain or off-chain settlement in step 4.

  1. 01Intake. Agent identity, intent, tool calls, vendor selection, and spend velocity captured pre-execution.
  2. 02Evaluate. Hard policy (allowlists, caps) combined with soft policy (prompt injection, catalog poisoning, intent deviation).
  3. 03Decide. Approve, block, or route to human review. Decision is signed cryptographically and written to the audit trail.
  4. 04Settle. Settle on Stripe, x402, Base, Solana, ERC-8004, or enterprise AP. Invoice auto-reconciles and syncs into your ERP.

## Supported rails

  • - Stripe
  • - x402
  • - Base
  • - Solana
  • - Tempo
  • - ERC-8004
  • - Enterprise AP

## Partners

Gordon integrates with and is partnered with the following:

## Endpoints

## Why now

Agents can spend. Now they can prove it. Existing payment infrastructure was built for human checkout — card networks, chargebacks, KYC — none of which model an agent's intent, scope, or attribution. Gordon binds a compliant wallet to the agent and audits every payment it makes, A2M or A2A.

## Meta

  • - Generated: 2026-06-18T13:34:16.693Z
  • - Source of truth: this URL
  • - License: see https://withgordon.ai/ for terms

$

View
HumanMachine