Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A Go CLI for Google's Universal Commerce Protocol — talk to UCP merchants over REST or MCP Trigger phrases: `use ucp`, `run ucp`, `check this UCP store`, `search UCP merchant`, `add to my UCP cart`, `prep a UCP checkout`.
.claude/skills/mvanhorn-pp-ucp/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 53% | 0% |
This skill drives the ucp-pp-cli binary. You must verify the CLI is installed before invoking any command from this skill. If it is missing, install it first:
$HOME/.local/bin on macOS/Linux and %LOCALAPPDATA%\Programs\PrintingPress\bin on Windows:bash npx -y @mvanhorn/printing-press-library install ucp --cli-only
ucp-pp-cli --version$PATH for the agent/runtime that will invoke this skill.If the npx install fails (no Node, offline, etc.), fall back to a direct Go install (requires Go 1.26.6 or newer):
bashgo install github.com/mvanhorn/printing-press-library/library/commerce/ucp/cmd/ucp-pp-cli@latest
If --version reports "command not found" after install, the runtime cannot see the binary directory on $PATH. Do not proceed with skill commands until verification succeeds.
v0.1 supports two interaction patterns:
| Command | Real merchants (e.g. checkout.coffeecircle.com) | Bundled mock (mock serve) | |---|---|---| | check <domain> | ✅ Any merchant publishing /.well-known/ucp | ✅ | | search / cart / checkout prep | ⚠️ Requires REST transport — most public merchants are MCP-only (deferred to v0.2) | ✅ Full end-to-end flow |
Known real merchants: checkout.coffeecircle.com (Shopify-hosted, MCP-only). v0.2 adds MCP transport client to unlock real-merchant transactions.
Use ucp-pp-cli when an agent needs to interact with UCP merchants directly: probe a manifest, build a multi-merchant cart, run a search across known UCP-supporting stores, or prep a checkout draft for an AP2 mandate. It is the natural local pair to an AP2 CLI for the search-and-buy half of an agentic shopping flow. Skip it when you only need ChatGPT-side checkout (use an ACP-flavored tool) or when you only need merchant-side validation (use awesomeucp/ucp-doctor).
These capabilities aren't available in any other tool for this API.
check — Fetch /.well-known/ucp for any domain and return a graded report covering schema validity, advertised transports, and capability coverage — works against any merchant publishing a UCP manifest, including MCP-only Shopify-hosted stores._Lets an agent screen a domain for UCP viability without standing up an SDK or hand-rolling manifest parsing._
bash ucp-pp-cli check checkout.coffeecircle.com --json
mock serve — Spawn a UCP-compliant reference merchant locally (pure-Go, no external runtime) so ucp check, search, cart, and checkout prep flows work end-to-end without a third-party UCP merchant or extra language toolchains._Lets an agent verify its UCP integration without coordinating with a live merchant or waiting for Google AI Mode approval._
bash ucp-pp-cli mock serve --port 8080
checkout — Operations on checkout
ucp-pp-cli checkout — POST /checkoutWhen you know what you want to do but not which command does it, ask the CLI directly:
bashucp-pp-cli which "<capability in your own words>"
which resolves a natural-language capability query to the best matching command from this CLI's curated feature index. Exit code 0 means at least one match; exit code 2 means no confident match — fall back to --help or use a narrower query.
bashucp-pp-cli mock serve --port 8080
Start the bundled pure-Go reference merchant; then run check 127.0.0.1:8080 from another shell.
bashucp-pp-cli search "coffee" --merchant 127.0.0.1:8080 --json
Fan out a query, normalize results, and pipe the first product into cart add then checkout prep.
bashucp-pp-cli check checkout.coffeecircle.com --json
Probe a Shopify-hosted UCP merchant. Returns a graded report even though the merchant advertises MCP-only transport (REST-only commands won't transact against it in v0.1).
UCP has no global API key. Each merchant declares its own auth shape; the CLI identifies itself via the UCP-Agent: profile="<url>" header and signs requests with a per-agent ECDSA-P256 key. Run ucp profile init once to generate your profile. For account-linked flows, ucp auth link <merchant> walks an OAuth 2.0 flow per merchant (v1: print-mode — agent pastes the redirect URL back).
Run ucp-pp-cli doctor to verify setup.
Add --agent to any command. Expands to: --json --compact --no-input --no-color --yes.
--select keeps a subset of fields. Dotted paths descend into nested structures; arrays traverse element-wise. Critical for keeping context small on verbose APIs:bash ucp-pp-cli checkout --agent --select id,name,status
--dry-run shows the request without sending--idempotent only when an already-existing create should count as successWhen you (or the agent) notice something off about this CLI, record it:
ucp-pp-cli feedback "the --since flag is inclusive but docs say exclusive"
ucp-pp-cli feedback --stdin < notes.txt
ucp-pp-cli feedback list --json --limit 10Entries are stored locally at ~/.ucp-pp-cli/feedback.jsonl. They are never POSTed unless UCP_FEEDBACK_ENDPOINT is set AND either --send is passed or UCP_FEEDBACK_AUTO_SEND=true. Default behavior is local-only.
Write what surprised you, not a bug report. Short, specific, one line: that is the part that compounds.
Every command accepts --deliver <sink>. The output goes to the named sink in addition to (or instead of) stdout, so agents can route command results without hand-piping. Three sinks are supported:
| Sink | Effect | |------|--------| | stdout | Default; write to stdout only | | file:<path> | Atomically write output to <path> (tmp + rename) | | webhook:<url> | POST the output body to the URL (application/json or application/x-ndjson when --compact) |
Unknown schemes are refused with a structured error naming the supported set. Webhook failures return non-zero and log the URL + HTTP status on stderr.
A profile is a saved set of flag values, reused across invocations. Use it when a scheduled agent calls the same command every run with the same configuration - HeyGen's "Beacon" pattern.
ucp-pp-cli profile save briefing --json
ucp-pp-cli --profile briefing checkout
ucp-pp-cli profile list --json
ucp-pp-cli profile show briefing
ucp-pp-cli profile delete briefing --yesExplicit flags always win over profile values; profile values win over defaults. agent-context lists all available profiles under available_profiles so introspecting agents discover them at runtime.
| Code | Meaning | |------|---------| | 0 | Success | | 2 | Usage error (wrong arguments) | | 3 | Resource not found | | 4 | Authentication required | | 5 | API error (upstream issue) | | 7 | Rate limited (wait and retry) | | 10 | Config error |
Parse $ARGUMENTS:
help, or --help → show ucp-pp-cli --help outputinstall → ends with mcp → MCP installation; otherwise → see Prerequisites above--agent)bash go install github.com/mvanhorn/printing-press-library/library/commerce/ucp/cmd/ucp-pp-mcp@latest
bash claude mcp add ucp-pp-mcp -- ucp-pp-mcp
claude mcp listwhich ucp-pp-cliIf not found, offer to install (see Prerequisites at the top of this skill).
--agent flag:bash ucp-pp-cli <command> [subcommand] [args] --agent
ucp-pp-cli <command> --help.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 46,228 | 65,735 | +42% | 1 | 1 | 0% | 2,804 | 2,594 | -7% | 0 | 0 | — |
case-02 | fail→fail | 37,802 | 35,905 | -5% | 1 | 1 | 0% | 1,302 | 2,715 | +109% | 0 | 0 | — |
case-03 | fail→fail | 5,758 | 5,664 | -2% | 1 | 1 | 0% | 933 | 2,652 | +184% | 0 | 0 | — |
case-08 | fail→pass | 43,256 | 2,726 | -94% | 1 | 1 | 0% | 1,972 | 2,785 | +41% | 0 | 0 | — |
case-04 | fail→fail | 42,234 | 4,999 | -88% | 1 | 1 | 0% | 2,096 | 3,114 | +49% | 0 | 0 | — |
case-05 | fail→pass | 45,060 | 4,444 | -90% | 1 | 1 | 0% | 2,161 | 3,038 | +41% | 0 | 0 | — |
case-06 | fail→fail | 16,257 | 4,727 | -71% | 1 | 1 | 0% | 2,234 | 3,120 | +40% | 0 | 0 | — |
case-07 | fail→pass | 13,536 | 2,994 | -78% | 1 | 1 | 0% | 2,164 | 2,806 | +30% | 0 | 0 | — |
case-09 | fail→pass | 15,606 | 3,110 | -80% | 1 | 1 | 0% | 2,727 | 2,799 | +3% | 0 | 0 | — |
case-10 | fail→pass | 12,208 | 2,790 | -77% | 1 | 1 | 0% | 1,799 | 2,753 | +53% | 0 | 0 | — |
case-11 | pass→pass | 4,534 | 1,884 | -58% | 1 | 1 | 0% | 684 | 2,612 | +282% | 0 | 0 | — |
case-12 | fail→pass | 11,337 | 3,012 | -73% | 1 | 1 | 0% | 1,867 | 2,760 | +48% | 0 | 0 | — |
case-13 | pass→pass | 8,498 | 2,087 | -75% | 1 | 1 | 0% | 1,240 | 2,585 | +108% | 0 | 0 | — |
case-14 | fail→pass | 13,854 | 2,599 | -81% | 1 | 1 | 0% | 1,949 | 2,707 | +39% | 0 | 0 | — |
case-15 | fail→pass | 12,550 | 3,313 | -74% | 1 | 1 | 0% | 2,078 | 2,868 | +38% | 0 | 0 | — |
case-16 | fail→pass | 9,009 | 2,452 | -73% | 1 | 1 | 0% | 1,481 | 2,671 | +80% | 0 | 0 | — |
case-17 | fail→pass | 11,270 | 3,094 | -73% | 1 | 1 | 0% | 1,746 | 2,795 | +60% | 0 | 0 | — |
case-18 | fail→pass | 13,368 | 2,394 | -82% | 1 | 1 | 0% | 1,988 | 2,701 | +36% | 0 | 0 | — |
case-19 | fail→pass | 11,931 | 3,423 | -71% | 1 | 1 | 0% | 1,906 | 2,827 | +48% | 0 | 0 | — |
case-20 | pass→pass | 7,667 | 3,319 | -57% | 1 | 1 | 0% | 1,159 | 2,830 | +144% | 0 | 0 | — |
case-21 | fail→fail | 9,127 | 6,067 | -34% | 1 | 1 | 0% | 1,441 | 3,322 | +131% | 0 | 0 | — |
case-22 | fail→fail | 9,757 | 2,787 | -71% | 1 | 1 | 0% | 1,490 | 2,750 | +85% | 0 | 0 | — |
case-23 | fail→fail | 8,874 | 2,760 | -69% | 1 | 1 | 0% | 1,334 | 2,772 | +108% | 0 | 0 | — |
case-24 | fail→pass | 16,166 | 4,734 | -71% | 1 | 1 | 0% | 2,514 | 3,061 | +22% | 0 | 0 | — |
case-25 | fail→fail | 13,237 | 8,458 | -36% | 1 | 1 | 0% | 2,604 | 2,606 | +0% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 25 cases were attempted, and 21 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +52 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.