Why We Built the UCP Registry
The UCP Registry started with a simple observation: the Universal Commerce Protocol was gaining traction, but there was no way to see the ecosystem. No directory of who had implemented it. No benchmarks for implementation quality. No API for agent developers to discover merchants programmatically. We built the Registry to fill that gap.
The Problem We Saw
When UCP launched in January 2026, adoption was immediate — but infrastructure wasn't. Within weeks, merchants across Shopify, WooCommerce, and custom platforms were spinning up endpoints, but there was no way to find them. If you wanted to discover UCP-enabled merchants, you had to:
- Manually check domains for
/.well-known/ucp.json - Ask around in Discord servers and GitHub discussions
- Maintain your own spreadsheet of known implementations
That worked in the first few weeks when there were a handful of early adopters. It wasn't going to scale as adoption accelerated through February and March.
For agent developers, the lack of a discovery layer was a blocker. You can't build a universal shopping assistant if you have to hardcode every merchant you want to support. The whole point of UCP is standardized discovery — but there was no standard place to discover who had adopted it.
What We Built
The UCP Registry is four things:
1. A Directory
The merchant directory is the public face of the ecosystem. Every UCP-enabled merchant gets a profile page showing their capabilities, response times, benchmark scores, platform, and AI bot policies. You can search, filter by category, and discover new merchants across dozens of verticals.
2. A Benchmark System
We don't just list merchants — we continuously test them. Every endpoint gets benchmarked across five dimensions:
- Speed — response time in milliseconds
- Data Quality — completeness and accuracy of product data
- Capability Coverage — how many UCP capabilities are implemented
- Manifest Compliance — adherence to the UCP spec
- Agent Accessibility — AI bot policies and endpoint availability
The result is a UCP Score (0-100) and a letter grade (A+ through F). This gives merchants a clear target and gives agents a quality signal for routing decisions.
The benchmarks are grounded in real agent behavior. Across 560+ tracked shopping sessions, we've observed that endpoint performance directly correlates with agent success:
| Metric | Ecosystem Average | Top Performers |
|---|---|---|
| Product search response | 460-980ms | <200ms |
| Checkout creation | 480ms | <350ms |
| Product listing error rate | 2% | <1% |
| Checkout error rate | 4-30% | <5% |
| Agent checkout success rate | 47% | 70%+ |
These aren't theoretical benchmarks — they're derived from 2,100+ real tool calls made by 16 different AI models. When we score a merchant, we're measuring how well they perform for the agents that actually use them. See the gold standard collection for top performers.
3. An API
The Registry API is how AI agents discover merchants programmatically. It supports filtering by:
- Status (verified, unverified)
- Platform (Shopify, WooCommerce, etc.)
- Capability (dev.ucp.shopping.catalog, dev.ucp.shopping.checkout, etc.)
- Category (apparel, electronics, beauty, etc.)
- UCP Score threshold
This is the layer that makes agent-powered product discovery work at scale.
4. A Monitoring System
We continuously crawl registered merchants to track:
- Endpoint uptime and availability
- Response time trends
- Capability changes
- Manifest updates
- AI bot policy changes
This means the Registry always reflects the current state of the ecosystem, not a stale snapshot.
Design Decisions
Open by Default
The Registry is designed to be as open as possible. The directory is public. The API is freely accessible. Merchant submissions are free. We believe the UCP ecosystem grows faster when information flows freely.
Quality Over Quantity
We verify every merchant endpoint before listing it as verified. A merchant with a broken or empty UCP manifest doesn't help agents or consumers. The benchmark system incentivizes merchants to implement well, not just implement.
Platform Agnostic
The Registry tracks merchants across every platform — Shopify, WooCommerce, Magento, BigCommerce, Salesforce Commerce Cloud, and custom platforms. UCP is a universal protocol, and the Registry reflects that universality.
What We've Learned
Adoption Follows Tooling
The biggest jumps in merchant adoption came when platforms added native UCP support. Making it a one-click enable rather than a manual implementation changed the trajectory. This reinforced our focus on tracking and promoting platform-level support.
Benchmarks Drive Improvement
Since launching the benchmark system, the average UCP Score across verified merchants has climbed from 62 to 76. Merchants pay attention to their scores and actively improve their implementations. Gamification works — even in B2B infrastructure.
The real-world impact is measurable. Checkout success rates across tracked sessions improved from 30% in early February to 79% on peak days in late March. Daily session volumes grew from single digits to 40-50 per day. And the total tracked GMV exceeded $108,000 across our sample — proof that quality improvements translate to real economic activity.
The Multi-Model Reality
One of the most surprising findings from our data: 16 different AI models from 7 providers are actively using UCP endpoints. Claude, GPT, Gemini, Llama, Grok, DeepSeek, and QWQ all show up in our session logs. This confirmed our design decision to keep UCP model-agnostic — the ecosystem is genuinely multi-vendor, not dominated by any single AI platform.
Performance varies significantly across models (success rates range from 25% to 71%), which informed our decision to test merchants against multiple models in our benchmark system. A merchant that works great with Claude but breaks with Gemini isn't truly UCP-compliant.
The Ecosystem is Bigger Than We Expected
What started as a directory for merchants quickly expanded to include platforms, plugins, tools, extensions, agents, and vendors. The UCP ecosystem isn't just merchants and agents — it's the entire infrastructure layer around them.
The scope of the ecosystem surprised us. We now track:
- 47+ unique stores with active agent traffic
- 7.5 tools per merchant endpoint on average — far richer than the basic search+checkout we initially designed for
- Multi-currency transactions across USD, EUR, GBP, and INR
- Over 24 million tokens processed in agent sessions — a proxy for the sheer volume of AI-commerce interaction happening on the protocol
What's Next
We're focused on several initiatives:
- Deeper benchmarking — testing actual product search quality, not just endpoint availability
- Real-time monitoring — minute-by-minute uptime tracking with alerting
- Ecosystem reports — regular state of the ecosystem analysis
- Developer tools — SDKs, testing utilities, and integration guides
The UCP ecosystem is growing fast. The numbers tell the story: thousands of merchants, dozens of platforms, millions of API queries per day. We're building the infrastructure to keep up.
Get Involved
- Merchants: Add your business to the ecosystem
- Developers: Build on the API and check out our agent building tutorial
- Everyone: Browse the directory to see what's possible
The Registry is the map of agentic commerce. We're drawing it as fast as the territory expands.
Explore the UCP Registry, read the API docs, or submit your store to join the ecosystem.
