# UCP Playground

> The independent test harness for agentic commerce. Point it at any store that publishes a Universal Commerce Protocol (UCP) profile: call the store's tools yourself over MCP, REST or WebMCP, run AI shopping agents against it, and replay every run call by call.

UCP Playground is independent: it is not run by a store, a platform or a model maker. Every model is tested on the same task, store and buyer context, and every result is a recorded run you can open and read. Inspect is deterministic (the same call gets the same answer); a run is probabilistic (the model chooses what to call and when to stop). Retail, Stays and Food test stores are available at flowers.ucpsandbox.com, stays.ucpsandbox.com and food.ucpsandbox.com; nothing is charged there.

## Main pages

- [Home](https://ucpplayground.com/): What the Playground does, with a recorded run you can replay.
- [Inspect](https://ucpplayground.com/inspect): Connect a store by domain, read its UCP profile and call its tools by hand, by step, by tool or as a buyer would. Signing in is required.
- [Runs](https://ucpplayground.com/runs): Start a run: give an agent a task at a store, pick the model, how it reaches the store, who it shops for and who the agent is. Signing in is required.
- [Model leaderboard](https://ucpplayground.com/leaderboard): Models ranked by how far their runs got (search, cart, checkout), by efficiency and by tokens.
- [Model directory](https://ucpplayground.com/models): One page per model with its shopping score, breakdown, known issues and token use.
- [Compare models](https://ucpplayground.com/models/compare): Two to four models side by side on the measures behind the shopping score.
- [Automated evals](https://ucpplayground.com/evals): Run a set of tasks across stores and models on a schedule and read the report.
- [The agent](https://ucpplayground.com/agent): Who our agent is when it calls a store: its UCP-Agent profile, how it signs, the capabilities it uses.
- [Features](https://ucpplayground.com/features): Everything the Playground records and shows.
- [How it works](https://ucpplayground.com/how-it-works): From inspecting a store to replaying a run.
- [Chrome extension](https://ucpplayground.com/extension): The shopping agent in your browser's side panel.

## Documentation

- [Docs](https://ucpplayground.com/docs): All guides.
- [What is UCP?](https://ucpplayground.com/docs/start-here/what-is-ucp): The Universal Commerce Protocol in brief.
- [Quick start](https://ucpplayground.com/docs/start-here/quick-start): From a store domain to your first run.
- [Understanding results](https://ucpplayground.com/docs/start-here/understanding-results): How a run ends and whose fault a failure was.
- [Inspect a store](https://ucpplayground.com/docs/test-a-store/inspector): Calling a store's tools by hand.
- [Agent runs](https://ucpplayground.com/docs/test-a-store/agent-runs): Starting and reading a run.
- [Choosing models](https://ucpplayground.com/docs/test-a-store/choosing-models): The models and their trade-offs.
- [UCP profile setup](https://ucpplayground.com/docs/be-tested/profile-setup): Publishing your UCP profile at `/.well-known/ucp`.
- [Testing your store](https://ucpplayground.com/docs/be-tested/testing-your-store): For store teams.
- [How scoring works](https://ucpplayground.com/docs/conformance/scoring): How the shopping score is computed.
- [API quickstart](https://ucpplayground.com/docs/automate/api-quickstart): Starting runs and evals from the API.
- [Common errors](https://ucpplayground.com/docs/troubleshooting/common-errors): What an error means and who fixes it.
- [Glossary](https://ucpplayground.com/docs/reference/glossary): The words used across the Playground.

## For agents

- [OpenAPI v1](https://ucpplayground.com/openapi/v1.yaml): The API description.
- [Agent profile](https://ucpplayground.com/.well-known/ucp-agent): The UCP platform identity our agent signs with.

## Reference

- [Changelog](https://ucpplayground.com/changelog): Release notes.
- [Privacy policy](https://ucpplayground.com/privacy)
- [Terms of service](https://ucpplayground.com/terms)

## Optional

- [UCP Checker](https://ucpchecker.com): Independent checks of UCP readiness across the store ecosystem.