Agent Commerce Observatory

Independent tracking of the Agentic Commerce Protocol ecosystem — Cycle 3 (2026-04-30)

Disclaimer: Independent tracking and analysis. Not affiliated with Stripe, Inc. or OpenAI. Not financial advice. Data sourced from public filings, developer documentation, and verified reporting.
Agent view: /?view=agent · Discovery: /.well-known/agent-discovery.json

Friction Gradient Index (FGI v0.1)

FGI = 1 − mean(Latency, Lock-In, Heat) · Higher = smoother rail · 🟢 ≥ 0.70 · 🟡 0.40-0.69 · 🔴 < 0.40

Rail	Latency	Lock-In	Heat	FGI	Delta	Notes
Visa Intelligent Commerce	0.10	0.05	0.10	0.92 🟢	UPGRADED	Interoperability bridge layer (Visa ICC, launched Apr 8 2026) — protocol-agnostic on-ramp across ACP, UCP, MPP, TAP. Card auth ~3s. No pre-funding. LatAm 5-market pilot + US closed beta live. Broader rollout targeted later 2026.
Mastercard Agent Pay	0.10	0.05	0.10	0.92 🟢	CONFIRMED	Santander EU first. Live transactions in ASEAN (Thailand KTC pilot, Apr 2026). Agent Suite Q2 2026.
Stripe ACP (card)	0.08	0.05	0.10	0.92 🟢	UPGRADED	Live infrastructure, seeded adoption. Platforms: Wizard, BigCommerce, Etsy, URBN, Shopify, Salesforce. Stripe ACS (hosted endpoint) is the primary onboarding accelerant. Merchant checkout retained by default after OpenAI discovery-first pivot.
Capital One EN-OAP	0.12	0.10	0.05	0.91 🟢	STALE	Virtual card gen + auth ~4s. Bank charter = near-zero heat.
Circle Nanopayments	0.07	0.25	0.20	0.83 🟢	NEW	Sub-second batched settlement. Gas-free sub-cent. Stablecoin reg pending.
UPI Agentic (Razorpay)	0.10	0.15	0.35	0.80 🟢	UPGRADED	Claude + ChatGPT + merchant apps. Reserve Pay caps = partial lock-in.
X Money	0.25	0.20	0.20	0.78 🟢	UPGRADED	6% APY, Visa debit, 44 MTLs. No agent capability yet.
Stripe x402 (USDC)	0.10	0.30	0.30	0.77 🟢	NEW	L2 on-chain ~3s. Full USDC pre-funding. Stripe also operates MPP (Machine Payments Protocol) for M2M alongside x402. CFPB + stablecoin reg ambiguity.
Coinbase x402/USDC	0.10	0.35	0.30	0.75 🟢	UPGRADED	V2 rollout, Cloudflare Foundation, Etherlink. $28K real daily volume.
Polymarket (agent)	0.10	0.40	0.40	0.70 🟢	UPGRADED	Polygon. Full pre-funding, self-custodial. Polystrat AI agent launched.
MoonPay Agents	0.50	0.30	0.20	0.67 🟡	CONFIRMED	Multi-hop lifecycle 30-120s. Non-custodial. Official Feb 24 launch.
Fetch.ai A2A	—	—	—	N/A ⏸️	DOWNGRADED	Insufficient production data. Visa integration unconfirmed on-chain.

Data Velocity Index (DVI v0.1)

DVI = DV-q × DV-r · DV-q = signal quantity per cycle · DV-r = label richness per signal · Measures how fast a rail learns from agent behavior.

Rail	DV-q (Quantity)	DV-r (Richness)	DVI	Notes
Polymarket (agent)	0.95	0.30	0.29	8,894 trades/run. Binary outcome labels only (yes/no).
Stripe ACP (card)	0.25	0.90	0.23	Pilot-stage volume, not production-wide scale. Rich labels: intent→product→settlement→satisfaction. Constrained by merchant-control friction.
Coinbase x402/USDC	0.40	0.25	0.10	50M cumulative but mostly noise. M2M settlement labels.
UPI Agentic (Razorpay)	0.20	0.70	0.14	Multi-platform pilots. Rich commerce labels (grocery, retail).
Visa Intelligent Commerce	0.15	0.85	0.13	Hundreds of controlled beta txns. Rich card-network labels but agent-specific data sparse.
Mastercard Agent Pay	0.10	0.85	0.09	Santander EU + ASEAN pilots (Thailand KTC). Same label depth as Visa.
Stripe x402 (USDC)	0.15	0.20	0.03	New rail. Stablecoin settlement labels only.
Circle Nanopayments	0.05	0.15	0.01	Testnet only. No production signals yet.
MoonPay Agents	0.10	0.35	0.04	Live but multi-hop lifecycle blurs signal attribution.
X Money	0.00	0.00	0.00	Zero agent data. No agent capability.
Capital One EN-OAP	0.05	0.50	0.03	STALE. Virtual card labels exist but no new agent signals.
Fetch.ai A2A	—	—	N/A	Deferred. Insufficient on-chain confirmation.

FGI × DVI Divergence Quadrant

🟢 High FGI + High DVI — Rideable AND Learning

No rail occupies this quadrant yet. This is the prize.

🔵 High FGI + Low DVI — Smooth but Blind

Visa (0.92 / 0.13) · Mastercard (0.92 / 0.09) · Capital One (0.91 / 0.03)

🟡 Low FGI + High DVI — Rough but Learning

Polymarket (0.70 / 0.29) — highest DVI of any rail

⚫ Low FGI + Low DVI — Dead Rail

X Money (0.78 / 0.00) · Fetch.ai (N/A / N/A)

Stripe membrane position: ACP FGI 0.92 · DVI 0.23 vs x402 FGI 0.77 · DVI 0.03 — osmotic spread Δ0.15 (under 0.25 threshold). The membrane holds.

Validation Alerts

[HIGH] ACP live but not widely adopted
ACP is in live production with seeded integrations (Wizard, BigCommerce, Etsy, URBN, Shopify, Salesforce). Broad merchant adoption remains early-stage. Key bottleneck: merchants resist delegating checkout — they retain transaction control and brand ownership. No confirmed public transaction-volume metrics.

[HIGH] Multi-protocol contest intensifying
Competing rails: ACP (Stripe/card), x402 + MPP (Stripe M2M), Coinbase x402 (USDC), Mastercard Agent Pay, UCP (rising governance layer — 10-member Tech Council including Stripe, Amazon, Meta, Shopify). Visa ICC operates as an interoperability bridge across protocols rather than a competing rail. Protocol choice — not just implementation quality — is now the dominant architectural variable.

[HIGH] Merchant-control friction is the adoption ceiling
OpenAI confirmed pivot away from native Instant Checkout to discovery + merchant-hosted checkout (Walmart app model). Users browse in-chat but rarely transact. Merchant risk aversion around delegating fulfillment, fraud, and KYC to agents remains the primary adoption blocker across all rails.

[HIGH] Volume transparency absent across all rails
No rail publishes confirmed agent transaction volume with a named primary source. x402 cites 50M+ cumulative but only ~$28K/day real commerce volume. All other rails report "pilot" or "launch partners" without activation counts. Any adoption-strength claim beyond seeded infrastructure is unsubstantiated.

[HIGH] Stripe cross-rail positioning
Stripe now sits across operation and governance layers of the field: ACP operator (card commerce), x402/MPP operator (USDC on-chain + M2M), and UCP Tech Council member (joined Apr 24 2026 alongside Amazon, Meta, Microsoft, Salesforce). This is structural positioning across competing protocol stacks — not convergence proof.

[HIGH] x402 adoption reality check
CoinDesk reports only ~$28K real daily volume despite 50M+ cumulative transactions. Infrastructure deployment dramatically outpacing actual commerce demand.

[MEDIUM] UPI platform divergence
Agentic UPI now on Claude AND ChatGPT. Platform-agnostic rail behavior validates "rail is the moat, not the agent" thesis.

[MEDIUM] x402 competitive threat
X-native builders frame x402 as losing to OpenClaw/Polymarket which embed payment natively rather than abstracting through protocol layer.

[LOW] Fetch.ai status downgrade
Grok classifies Fetch.ai A2A as ANNOUNCED not LIVE-PRODUCTION. On-chain confirmation pending.

Ghost Rail Watch

Rails with volume/capacity < 0.001 for two consecutive cycles.

Rail	Cycles Stale	Capacity	Agent Volume	Flag
📂 Pix (Brazil)	2	8B monthly human txns	Zero	—
📂 UnionPay	2	MCP announced	Zero	—
📂 NEAR AI	2	Unknown	Zero	—
📂 x402	1	50M+ cumulative	$28K/day real	WATCH — gamed metrics

Convergence Signal

Protocol competition is accelerating faster than broad merchant adoption

Traditional cluster: 0.91-0.92 Converging toward production through controlled pilots within regulated frameworks. Merchant-control friction — not infrastructure readiness — is now the primary adoption ceiling.

Crypto cluster: 0.70-0.77 Production-grade infrastructure but negligible real commerce volume

Membrane holder: Stripe (operator: ACP, x402/MPP; governance: UCP Tech Council member) · Osmotic spread: Δ0.15

Grok convergence read: Protocol infrastructure is maturing faster than broad merchant adoption, and competition between rails is now the dominant shaping force.

Next Gate Nodes

Visa Intelligent Commerce → broader production rollout (targeted later 2026)
X Money public rollout (passed — still EXTERNAL BETA, agent capability not launched)
Merchant-control friction resolves OR agent checkout conversion improves materially
First named rail publishes confirmed agent transaction volume with primary source
CFPB fiduciary determination
BoE/MAS Q2 consultations
Circle nanopayments mainnet
UCP/Google status confirmed (currently unverified in production)

Security wildcard: OpenClaw 135K stars, 21K exposed instances. High-profile consumer loss could trigger regulatory freeze across both tracks.

Methodology

Friction Gradient Index (FGI)

Three-pillar seismograph measuring rail rideability today.

Latency: Time from agent intent to settlement confirmation (t_settle / 30s, clamped to 1)
Lock-In: Collateral or reserve requirements to transact (locked_capital / tx_amount, clamped to 1)
Regulatory Heat: Proximity to enforcement action (enforcement_score / 10)

Data Velocity Index (DVI)

Companion metric measuring how fast a rail learns from agent behavior.

DV-q (Quantity): Agent-specific labeled signals per cycle
DV-r (Richness): Label depth per signal. Binary outcomes = 0.1-0.3. Intent→settlement = 0.5-0.7. Full lifecycle = 0.8-1.0
DVI = DV-q × DV-r (normalized 0-1)

Machine-readable methodology: /.well-known/fgi-methodology.json

Integration Hub

Discovery-first integration guides for merchant teams navigating a multi-protocol reality. ACP (card), x402 (USDC), and competing rail standards are all live. The dominant structural pattern is merchant-owned checkout handoff: agents drive discovery, merchants retain transaction control. Stripe ACS (hosted ACP endpoint) is the primary onboarding accelerant for Stripe-native stacks.

Product Feed Quickstart
Prepare a merchant catalog for AI-native product discovery in under 30 minutes.
JSON: /guides/product-feed-quickstart.json
Shopify Catalog Integration
Understand what Shopify auto-sync covers and what merchants still must provide manually.
JSON: /guides/shopify-catalog-integration.json
Merchant-Hosted Checkout Path
Decision guide for discovery-first integration where merchants retain checkout control.
JSON: /guides/merchant-hosted-checkout-path.json

Guide index: /guides/ · Machine index: /guides/index.json · Feed: /feed.json