Skip to main content

Marketing Platform

The Boat
Berley attracts. The boat catches. Build both.
Platform means bait, gear, and technique — matched to the fish and the water, not copied from a trend report.
AI generates
Content, personalization, and agent workflows at machine scale.
Blockchain proves
Attribution, loyalty, and intent become verifiable on-chain.
Crypto aligns
Token incentives align creator, contributor, and customer interests.

The agentic-commerce stack is not a single product — it is seven layers. Each layer has a job. The combination matters. Wrong bait for the fish, wrong technique for the water — no catch.

The frame is ABCD: AI generates → Blockchain proves → Crypto aligns → DePIN distributes.

Layer 1: Identity and Authorization​

Before an agent can act on a buyer's behalf, it needs provable delegation from a human principal.

The problem this solves — when a human taps a card, the tap IS consent. When an agent transacts, there is no tap. Identity and auth layers supply the cryptographic equivalent.

  • Verifiable Intent — open-sourced by Mastercard and Google; a delegation chain binding agent actions to human-approved scope. See full specification.
  • Visa Trusted Agent Protocol — Visa's framework for agent-authorized payments within defined limits.
  • Embedded-wallet identity — Privy and Dynamic offer embedded wallets that let users own keys without managing seed phrases, enabling seamless web2 → web3 identity bridges.

The question this layer answers: who authorized the agent, and what are they allowed to do?

Allowlist Architecture​

The pre-authorization work a brand does before an agent can transact with them. This is the discipline the identity layer enables but does not perform: earning a place on the L2 Verifiable Intent merchant-allowlist that governs whether an agent is authorized to transact at all.

The allowlist constraint is set by the human principal — once, before any purchase cycle. If a service is absent, it is structurally excluded from every autonomous transaction the agent makes. This is the layer between identity proof and payment execution that most protocol documentation skips.

What allowlist architecture involves:

  • Publishing a VI-compatible merchant profile: machine-readable declarations of constraint-satisfaction capabilities (accepted budget caps, category fit, approval thresholds)
  • Registering on the UCP merchant registry at ucp.dev — the structural equivalent of submitting a sitemap to Google Search; the Universal Cart launching in Google Search and Gemini in summer 2026 [source: Google I/O 2026] discovers merchants through this registry
  • Maintaining /.well-known/agent.json with constraint boundaries a principal can verify before encoding the allowlist entry
  • Building principal-facing content that proves constraint-satisfaction, not product features — the principal never transacts directly; they need proof that their agent's scope will be respected

UCP as the sleeper protocol: UCP launched January 2026 with 20+ major retail and payment partners (Shopify, Target, Walmart, Visa, Mastercard, Adyen, Stripe) [source: Google Developer Blog, January 2026]. Registering in the UCP merchant registry is not optional for services targeting agent-mediated commerce through Google surfaces — it is the table-stakes equivalent of search indexation.

The question this layer answers: have you done the work that earns a principal's trust to encode you into their agent's authorized scope?

Layer 2: Payment Execution​

Three protocol families are competing to become the default rails for agent transactions. The standards war detail is at the commerce hub — this is the marketing lens.

x402 — per-call micropayments, developer-first, Coinbase + Google + Linux Foundation. 165M+ cumulative transactions as of April 2026 [source: Coinbase Developer Platform, April 2026]; daily volume dropped 92% from December 2025 peak as early bot-driven volume transitions to organic demand [source: CoinDesk, March 2026]. Strongest current fit: API monetization and DePIN data feeds. Marketing value: x402 is an attribution rail — every on-chain transaction is a verified conversion event, closing the attribution gap that web2 analytics cannot close.

ACP (Agentic Commerce Protocol) — OpenAI + Stripe; agent-to-merchant checkout. PayPal adopting for 2026. Best when the buyer is an agent and the seller is a merchant with existing Stripe infrastructure.

AP2 (Agent Payments Protocol) — Google; enterprise authorization via Verifiable Delegated Credentials. 60+ partners including Mastercard, AmEx, Coinbase. Sui as launch partner. Best for enterprise workflows where audit trails and authorization boundaries matter.

UCP (Universal Commerce Protocol) — Google's capability-discovery and cart standard; helps agents find services, verify capabilities, and initiate checkout. 20+ founding partners including Shopify, Walmart, Visa, Mastercard, Adyen, Stripe [source: Google Developer Blog, January 2026]. The Universal Cart launches inside Google Search and Gemini in summer 2026. Registration at the UCP merchant registry (ucp.dev) is the direct path to being discoverable in that surface — see Allowlist Architecture above.

The question this layer answers: how does value move from buyer agent to seller, with provable authorization?

Layer 3: Settlement​

Payment protocols describe negotiation. Settlement is where value actually moves.

  • Base — Ethereum L2, Coinbase-backed; home of x402's primary implementation. Low fees, fast finality, large developer ecosystem.
  • Sui — ~390ms finality, Move language safety, Google's AP2 launch partner. [Source: Sui docs]. The team that built Meta's Diem payment infrastructure rebuilt it permissionlessly here.
  • Solana — ~400ms finality [source: Solana docs], high throughput, large DeFi and consumer ecosystem.
  • Stablecoins — USDC and USDT settle instantly, borderlessly, without traditional banking rails. The instrument agents use when fiat settlement tempo is too slow.

Traditional rails cannot settle at machine tempo across borders at sub-cent amounts. Stablecoins on high-throughput chains close the gap between intent and finality.

Layer 4: Discovery​

If agents can't find you, you don't exist. Discovery in agent-native markets is not search ranking — it is protocol registration and machine-readable context.

  • MCP servers (Model Context Protocol) — expose your capabilities as tools that AI agents can call directly. Anthropic → Linux Foundation. The HTTP of agent-to-tool access.
  • A2A (Agent-to-Agent Protocol) — Google's coordination layer for agent-to-agent negotiation before a transaction. The TCP of agent connections.
  • /.well-known/agent.json — a standardized endpoint declaring what your service does, what it costs, and which payment protocols it accepts. The DNS record of agent commerce.
  • agenticcommerce.dev registry — the emerging directory for agent-accessible services. agenticcommerce.dev
  • Agent Intent Optimisation (AIO) — being cited in AI answer surfaces (ChatGPT, Perplexity, Claude, Gemini) rather than just ranked in search. The berley trail for agents. Machine-readable docs, structured schema, and citation-worthy content are the new SEO.

The question this layer answers: how does an agent discover that you exist, what you offer, and how to pay?

Layer 5: Attribution and Analytics​

In agent-native markets, traditional last-click attribution breaks down. An AI agent may research across 20 sources, synthesize a recommendation, and trigger a purchase — none of which looks like a session in Google Analytics.

Onchain analytics — Dune and Nansen provide wallet-level behavioral data, on-chain transaction flows, and protocol-level metrics. The source of truth when settlement happens on a chain.

Crypto-native attribution — Spindl and Formo track wallet-based funnels: who connected a wallet, what they did before and after, which campaigns drove onchain actions. Built for web3 where cookies are irrelevant.

Web2 leg analytics — product analytics tools remain essential for the web2 portion of the funnel: session behavior, content engagement, conversion from human-driven visits. These complement, not replace, onchain attribution.

Signal data enrichment — Clay, Clearbit, and BuiltWith enrich prospect records with technographic and firmographic data. Know which protocols a target company is already using before you reach out.

The question this layer answers: what actually caused the conversion, and which berley content is worth laying more of?

Layer 6: Content and Distribution​

The berley trail itself — the machine-readable, human-resonant content that agents cite and buyers trust.

Machine-readable knowledge base — structured documentation, schema-marked content, and well-organized knowledge base pages are the primary discovery surface for AI agents. If your docs are dense tables and PDFs, agents skip you.

Newsletter and long-form platforms — Beehiiv and Substack for owned distribution to human subscribers. Email remains the highest-leverage direct channel for human readers. Build it before you need it.

Farcaster Frames — interactive, composable content units that live natively on the Farcaster social protocol. Enable transactions (mint, subscribe, pay) directly from a social post without leaving the feed. The onchain equivalent of a landing page embedded in a tweet.

Social listening and audience intelligence — SparkToro maps where your audience actually spends time, not where you assume they are. Kaito tracks crypto-native mindshare and narrative momentum.

Content and video production tools — AI-assisted writing, short-form video (Descript, Riverside, VEED), and graphic production (Canva) at the volume a modern berley trail demands.

The question this layer answers: how does your content reach humans and agents, in formats they can both act on?

Layer 7: Onchain Loyalty​

Loyalty in agent-native markets is programmable. Rewards are verifiable. Membership is portable.

  • ThirdWeb Loyalty — token-gated loyalty programs, onchain reward distribution, composable with other protocol layers.
  • Unlock Protocol — NFT-based memberships and subscriptions. A user's membership token moves with them across platforms; the brand doesn't own the relationship, the user does — which paradoxically increases trust.
  • POAPs (Proof of Attendance Protocol) — onchain attendance records for events, milestones, and community participation. Collectible proof of engagement that accumulates into a verifiable reputation.
  • Token-gated community — Discord, Circle, and Telegram communities gated by token ownership. Contributors who hold tokens are aligned, not just subscribed.

The question this layer answers: how do you reward loyalty in a way the customer owns, can prove, and carries with them?

How to Choose Your Layer​

Start from the job to be done, not the technology.

If your buyers are primarily human today — invest heavily in Layers 4 (discovery/AIO), 6 (content), and 5 (analytics). These compound fastest with human-driven traffic.

If you're building for agents as buyers — Layer 2 (payment execution) and Layer 1 (identity/auth) are the enabling infrastructure. Without them, agents can't transact with you even if they find you.

If you're building a community — Layer 7 (onchain loyalty) combined with Layer 6 (owned distribution) creates the flywheel. Token alignment keeps contributors engaged when novelty fades.

If you're a data-rich business — Layer 5 (attribution) is your competitive advantage. Understanding which berley works is the compounding edge that ad spend alone cannot buy.

The stack grows over time. Most businesses start with Layer 6 (content), add Layer 4 (discovery/AIO), then add Layer 5 (attribution) once they have enough signal to optimize. Layers 1–3 become relevant when you serve agent buyers or want programmatic, permissionless revenue.

Questions​

Which layer of the stack is the bottleneck on your berley trail right now?

  • If AI agents become a significant fraction of discovery in the next 24 months, which layer are you least prepared for — and what is the cost of waiting?
  • Where is AI replacing human judgment in your marketing stack versus amplifying it — and where should the boundary be?
  • Which tool in your current stack generates the most value per dollar — and how would you prove it onchain?