The Settlement Layer for Agentic Commerce

Sometime in the next three years, a procurement agent in Detroit will negotiate a price with a parts supplier in Monterrey, confirm delivery terms, and settle payment — all without a human touching a keyboard. That transaction will need to clear in seconds, not days. It will need to move across borders, currencies, and regulatory regimes. And it will need to do so under programmable rules set by the businesses on both sides.

Nothing that exists today can do this.

The infrastructure being built for agentic payments — Natural in the U.S., Stripe's Agentic Commerce Suite, Visa's Trusted Agent Protocol — is designed for domestic transactions on fiat rails. But the defining characteristic of B2B commerce is that it crosses borders. $872.83 billion crossed the U.S.-Mexico border alone in 2025 — a record, and the highest bilateral trade volume the United States has ever recorded with any nation. Globally, corporate cross-border payments total $23.5 trillion annually.

AI agents are not going to respect national boundaries. The payment infrastructure they run on shouldn't either.

8os is building the settlement layer for cross-border agentic commerce: stablecoin-native payment infrastructure for AI agents, with local on/off-ramps to domestic rails — starting with the largest trade corridor in the Americas.

The agentic payments space is attracting serious capital. Natural raised $9.8M to build agent wallets and payment APIs over U.S. fiat rails. Catena Labs raised $18M to build an AI-native financial institution on stablecoins. Skyfire partnered with Visa. Stripe co-developed the Agentic Commerce Protocol with OpenAI. Nekuda raised $5M backed by both Visa Ventures and Amex Ventures.

All of them are solving the same problem: how does an agent pay for something domestically in the United States?

None of them are solving the harder, larger problem: how does an agent settle a cross-border B2B transaction in real time?

Today, when a business in the U.S. pays a supplier in Mexico, the payment travels through SWIFT — a 50-year-old messaging network that takes 3-5 business days, charges $25-45 per wire, requires correspondent banking relationships, and provides minimal transparency into where the money is at any given moment. The average cost of sending $200 cross-border is still 6.2%, far above the G20's target of 3%.

For humans, this friction is annoying. For autonomous agents operating in real time, it is a hard blocker. An agent cannot negotiate, confirm, and settle a transaction if the settlement leg takes five days. The entire value proposition of agentic commerce — speed, autonomy, always-on operation — breaks down at the border.

This is not a niche problem. B2B transactions account for 72.6% of all cross-border payment volume. B2B e-commerce payments alone are projected to reach $21.9 trillion by 2030. The cross-border payments market generates $193 billion in annual revenue, with the average transaction costing over $20 in fees. McKinsey projects that agentic commerce could drive $3-5 trillion in global GMV by 2030. Even capturing the settlement infrastructure for a fraction of the cross-border agentic B2B segment represents a market measured in tens of billions.

The companies building domestic agent payment rails are solving a real problem. But they are building for the geography where agents happen to operate today — the U.S. — not for the topology of how agents will operate tomorrow: across borders, across currencies, across regulatory regimes, in real time.

There is a payment rail that already operates globally, settles in seconds, costs fractions of a cent, is programmable, and doesn't require correspondent banking relationships. It's not a theoretical technology or a protocol under development. It processes over $27.5 trillion in annual transaction volume — more than Visa and Mastercard combined.

Stablecoins.

The argument for stablecoins as the settlement layer for cross-border agentic commerce is not ideological. It is purely functional:

Settlement speed. SWIFT: 3-5 business days. ACH: 1-3 business days. Stablecoins: seconds. For agents operating autonomously and transacting in real time, settlement speed is not a nice-to-have — it is a prerequisite. An agent that negotiates a freight rate and needs to lock in the price cannot wait for a wire to clear.

Programmability. Stablecoin transfers are programmable by default — they can be conditioned on events, gated by smart contracts, and composed with other on-chain operations. This maps perfectly to the agentic paradigm where policy engines need to enforce rules at the transaction level: spending limits, approved counterparties, conditional releases, escrow-and-release on delivery confirmation.

Borderless by default. USDC doesn't care if the sender is in Detroit and the receiver is in Monterrey. There is no correspondent banking chain, no intermediary fees, no currency conversion delay. The settlement is atomic and final.

Cost. The average cross-border B2B payment costs over $20 in fees. A USDC transfer on Base or Solana costs under $0.01. Even after on/off-ramp costs, the total transaction cost is 80-90% lower than traditional rails.

Institutional readiness. This is no longer a fringe argument. Stripe acquired Bridge — a stablecoin infrastructure company — for $1.1 billion, the largest acquisition in crypto history. Juniper Research identified stablecoins as key to unlocking B2B payments efficiency, stating that they bypass many traditional cross-border issues. Circle's USDC is regulated and transparent. The U.S. is actively advancing stablecoin legislation. Stablecoins have crossed the threshold from speculative instrument to financial infrastructure.

The critical missing piece is not the stablecoin itself — it is the infrastructure layer that connects the stablecoin to the local financial system at each end of the transaction. When an agent in the U.S. sends USDC to pay a supplier in Mexico, someone needs to receive that USDC and deposit Mexican pesos into the supplier's bank account via SPEI within seconds. Someone needs to handle the KYB verification, the CFDI invoice generation, the policy enforcement, and the audit trail.

That infrastructure layer is what 8os builds.

The choice of starting corridor is not incidental — it is the strategic thesis.

$872.83 billion in bilateral trade. The U.S.-Mexico trade corridor is the largest in the Americas and the single largest bilateral trade relationship the United States has with any nation in the world. Mexico has been the U.S.'s top trading partner for three consecutive years, surpassing both Canada and China.

Nearshoring is accelerating, not slowing. The effective U.S. tariff rate on Mexican goods is 4.18%, compared to 30.93% on Chinese imports. Mexico is on track to surpass China as the largest source of advanced technology product imports to the U.S. Manufacturing supply chains are actively relocating from Asia to Mexico. Every factory that moves increases cross-border B2B payment volume.

Deep supply chain integration. This is not arms-length trade. U.S. and Mexican manufacturing are deeply integrated — components cross the border multiple times during production. Port Laredo alone handles $354 billion in annual two-way trade. The automotive, electronics, and industrial sectors operate as a single cross-border production network.

SPEI is world-class infrastructure. Mexico's domestic payment system, SPEI, is one of the most advanced real-time payment systems in the world — processing over 1.2 billion transactions annually with settlement in seconds, 24/7. This means the "last mile" — converting stablecoins to local currency and delivering to the recipient's bank account — can happen in real time. The on/off-ramp is not the bottleneck in Mexico; the cross-border leg is.

The corridor is underserved by agentic infrastructure. Natural, Skyfire, Nekuda, and Stripe's agentic tools are built for the U.S. domestic market. No one is building agent payment infrastructure specifically for the cross-border B2B flows that define the U.S.-Mexico economic relationship.

The template is replicable. Every corridor we enter after Mexico follows the same playbook: stablecoin settlement + local on/off-ramp to domestic real-time payment rail. Brazil (PIX), Europe (SEPA Instant), India (UPI), Southeast Asia (PromptPay, QRIS) — each has a high-quality domestic rail that becomes the exit point. Our architecture is corridor-agnostic by design; only the local integration changes.

8os is the financial infrastructure layer that enables AI agents to move money across borders instantly, compliantly, and under programmable business rules.

Core Architecture

Organization → User → Agent Identity. Every entity in the system is part of a hierarchy. A business completes KYB verification and becomes an Organization. Users within that organization manage the account. Agents — autonomous software programs acting on behalf of the organization — receive delegated credentials with scoped permissions. Each level of the hierarchy inherits and can further restrict the policies above it.

Policy Engine. Every transaction passes through a multi-level authorization framework before execution. Organization-level policies set global constraints (maximum daily volume, approved jurisdictions, required compliance checks). Agent-level policies add agent-specific rules (per-transaction limits, approved counterparties, spending categories). Transaction-level controls enable real-time conditional logic (human-in-the-loop approval above certain thresholds, escrow requirements for new counterparties). These policies are deterministic guardrails — they operate independently of the agent's LLM reasoning, ensuring that even a hallucinating agent cannot execute a transaction that violates business rules.

Rail Abstraction. The payment execution layer abstracts the underlying rail. When an agent calls ochenta.payments.create(), the system selects the optimal rail based on the transaction parameters: stablecoin transfer for cross-border settlement, SPEI for domestic Mexican disbursement, and future integrations for additional corridors and rails. The developer does not need to know how the settlement happens — only that it happens, fast and compliantly.

Audit & Observability. Every transaction generates a complete audit trail: which agent initiated it, under what policy it was authorized, which rail was selected, the settlement status, and the compliance checks performed. For non-deterministic actors like AI agents, this observability layer is critical — it enables businesses to monitor, backtest, and prove the correctness of their agents' financial decisions.

Products

Wallet. Multi-currency accounts where agents hold, receive, and send funds. A single balance interoperates across stablecoins and fiat, with controllable sub-wallets per agent and just-in-time funding to minimize exposure. USDC, and fiat (MXN, USD) in one unified balance.

Pay. A single API call to send money to anyone, anywhere — cross-border or domestic. 8os handles rail selection, compliance, FX conversion, and settlement. For the U.S.-Mexico corridor: USDC in, MXN out via SPEI, settled in seconds.

Collect. Enables agents to receive payments from customers, vendors, and other agents — through payment links, direct transfers, or inter-wallet movements.

Bill. Agentic invoicing with automatic CFDI generation for Mexican tax compliance. When an agent completes a service, it can generate a legally compliant invoice, send it, and collect payment — programmatically. No human accountant required. This product has no equivalent anywhere in the world.

Transfer. Internal fund movements between wallets, organizations, and external accounts.

Bridge. The on/off-ramp engine: USDC ↔ MXN conversion with real-time settlement via SPEI. This is the core infrastructure primitive that makes everything else possible.

Developer Experience

The API is designed to be implemented in an afternoon. SDKs in Python and TypeScript. Typed responses. Idempotency keys. Prefixed IDs (org_, agt_, pay_, wal_). Comprehensive documentation. Webhook support for asynchronous events. MCP server integration for direct agent-to-API communication.

import ochenta

client = ochenta.Client(api_key="sk_live_...")

# An agent pays a Mexican supplier, cross-border, in one call
payment = client.payments.create(
    amount=50000,
    currency="MXN",
    from_agent="agt_procurement_01",
    to_party="pty_supplier_monterrey",
    rail="auto",  # 8os selects optimal rail
    memo="PO-2026-0847 — brake components",
    idempotency_key="idm_abc123"
)
# Settlement: USDC → MXN via SPEI. Seconds, not days.

The agentic payments market is fragmenting into four distinct segments. 8os occupies the one that no one else is building for.

Domestic fiat agent infrastructure (Natural, Stripe). Natural is building the best domestic agent payment stack in the U.S. — wallets, pay, collect, billing, identity, all over ACH and card rails. Stripe is extending its existing merchant infrastructure with the Agentic Commerce Suite and ACP. Both are fiat-first, U.S.-first, and consumer/domestic-focused. When their customers need cross-border B2B settlement, they will need exactly what 8os builds.

Stablecoin-native financial institutions (Catena Labs). Catena, co-founded by the Circle co-founder, is building the first AI-native bank on stablecoin rails — seeking actual banking licenses, focused on becoming a regulated financial institution. Their scope is broader (full banking) but their product is narrower (they are not building B2B payment orchestration or local on/off-ramps).

Consumer checkout for agents (Nekuda, Skyfire). Nekuda enables agents to complete purchases on existing e-commerce sites via browser automation. Skyfire provides a payment and identity network for agents consuming digital services. Both focus on agents buying things (C2B), not on agents settling B2B invoices across borders.

Incumbent extensions (Visa, Mastercard, PayPal). Visa's Trusted Agent Protocol, Mastercard's Agent Pay, and PayPal's Agent Ready are all extensions of consumer payment networks into the agent paradigm. They serve the existing card-based ecosystem, not the new cross-border B2B infrastructure requirement.

8os: Cross-border B2B agent settlement infrastructure. We are stablecoin-native (like Catena) but focused on B2B payment orchestration (like Natural), with local on/off-ramps that no one else is building. Our competitive position is unique because we don't compete head-on with any single player — we occupy the intersection that none of them cover.

The question is not whether agents will need to make cross-border B2B payments. The question is who builds the infrastructure when they do.

8os monetizes at three layers:

Transaction fees on payment volume. Every payment that flows through 8os generates revenue — a percentage of transaction value plus a fixed fee, similar to Stripe's model. Cross-border transactions command higher margins than domestic payments due to the complexity involved.

FX conversion margin. Every cross-border transaction involves a currency conversion (USDC ↔ MXN, USDC ↔ BRL, etc.). 8os captures a spread on this conversion, transparent and competitive relative to traditional banking alternatives but significantly more efficient.

Value-added services. Premium features including advanced policy engine configurations, enhanced observability and analytics, compliance automation (CFDI generation, SAT reporting), and SLA-backed settlement guarantees.

As the network grows, unit economics improve: more recipients on-platform means fewer off-ramp transactions, creating a flywheel where increased adoption reduces costs and increases margins.

Four forces are converging simultaneously:

Agentic AI has reached production readiness. In 2024, agents were demos. In 2025, they entered production. OpenAI, Anthropic, Google, and hundreds of startups are deploying agents in real business workflows. Salesforce reported that AI agents influenced $67 billion in global Cyber Week 2025 sales. Visa observed a 1,200% year-over-year increase in AI agent traffic to retail sites. The demand side is materializing.

Stablecoins have achieved institutional legitimacy. Stripe's $1.1B acquisition of Bridge. USDC on major institutional platforms. Active U.S. legislation on stablecoin regulation. Juniper Research calling stablecoins the key to unlocking B2B payment efficiency. The supply side — the rails — is ready.

Nearshoring is restructuring North American supply chains. The U.S.-Mexico trade relationship is not just large — it is accelerating. Companies are actively moving production from Asia to Mexico, creating new cross-border payment flows that didn't exist five years ago. Mexico is becoming the manufacturing backbone of the Americas.

No one has built this yet. Despite $50M+ deployed across agentic payment startups, zero dollars have been invested in cross-border B2B agent settlement infrastructure. The gap is not small — it is total. The window to establish the defining infrastructure in this category is 12-18 months before incumbents (Stripe, Wise, Bridge) or well-funded startups (Natural, Catena) expand into cross-border.

Phase 1 — Wedge (Now → Month 6). Launch the USDC ↔ MXN on/off-ramp with SPEI settlement. Onboard first enterprise customers — companies with existing cross-border B2B payment flows between the U.S. and Mexico. Generate revenue. Prove that stablecoin settlement + local rail disbursement works in production.

Phase 2 — Agent-Native (Month 6 → Month 12). Launch agent identity, policy engine, controlled wallets, and the full payment API. Enable AI agents to autonomously initiate, authorize, and settle cross-border payments within programmable guardrails. Launch agentic billing with automatic CFDI generation.

Phase 3 — Corridor Expansion (Month 12 → Month 24). Add the U.S.-Brazil corridor (PIX on/off-ramp). Begin SEPA integration for U.S.-Europe. Each new corridor multiplies network value and demonstrates that the architecture is genuinely corridor-agnostic.

Phase 4 — Settlement Network (Month 24+). Become the default settlement layer for agentic cross-border commerce. Enable agent-to-agent payments across any corridor. Launch credit products, advanced treasury management, and multi-currency optimization.

We are building 8os from Mexico City, with deep knowledge of Mexican financial infrastructure, regulatory environments, and the specific operational realities of the U.S.-Mexico corridor.

Braulio — Founder & CEO. Builder at the intersection of payments infrastructure and AI. Previously built and scaled digital businesses across multiple markets. Student at Tec de Monterrey and Universidad Panamericana, with direct access to Mexico's most significant business networks. Obsessed with the conviction that the largest companies of the next decade will be built at the intersection of AI and financial infrastructure.

We are raising a seed round to execute Phase 1 and Phase 2: build the USDC ↔ MXN on/off-ramp, launch the agent payment API, onboard our first enterprise customers, and prove that stablecoin-native cross-border settlement is the right architecture for agentic B2B commerce.

Today, $23.5 trillion in cross-border B2B payments flows through infrastructure that was designed for humans sending messages to other humans. SWIFT was revolutionary in 1973. It is a bottleneck in 2026.

The next generation of cross-border commerce will be executed by autonomous agents that negotiate, transact, and settle in real time — 24 hours a day, 7 days a week, across every border. Those agents will need a settlement layer that matches their speed, their autonomy, and their global reach.

Stablecoins provide the rail. Local payment systems provide the last mile. What's missing is the intelligence layer in between — the identity, the authorization, the policy enforcement, the compliance, the observability, and the orchestration that turns a raw stablecoin transfer into a trusted, auditable, regulatorily compliant business payment.

That layer is 8os.

We are not building the next Stripe. We are not building the Natural of Latin America. We are building the settlement infrastructure for a world where machines trade with machines across borders — the SWIFT of autonomous commerce, native to the rails that agents deserve.

The window is open. The corridor is ready. The technology exists. The question is who builds it first.