UCP Playground

Developer tool for building and debugging AI shopping agents against live UCP-enabled stores (Shopify, WooCommerce, and more). - Debug console — Full request/response logging with timing (ms), JSON payload inspection, and transport-level tracing across MCP, REST, and Embedded protocols - Schema quality analyzer — Scores tool schemas A-F, catches missing items definitions, absent required fields, untyped parameters, and undescribed properties before they break your agent - Tool inspector — Browse discovered MCP tools, inspect input schemas, invoke manually with a JSON editor, and examine raw responses - Multi-model benchmarking — Run up to 5 LLMs side by side against the same store and prompt, compare token usage, tool call count, duration, and checkout success rate - Session replay with backend event stream — Step through every model call, tool invocation, and response with nanosecond-precision timing. Toggle between chat view and a dark terminal event log - Store audit engine — Automated 4-phase UCP capability scan: manifest discovery, MCP initialize, search probe, and capability detection. Scores performance (TTFB), data quality, and capability coverage - Transport fallback debugging — Watch MCP-to-REST fallback in real time, per-tool. See which calls succeeded via MCP and which fell back, with error codes (-32601, HTTP 4xx/5xx) - Token usage — Prompt/completion token breakdown per turn, per session, and per model across the leaderboard - OAuth flow inspector — Visualize authorization code + PKCE flows, track token lifecycle, inspect scopes and refresh state - Embeddable session replays — Share debug sessions via public URLs or embed as iframes for bug reports and documentation