Roadmap
UCP compliance is the entry point. Here's what makes PennyLane a platform, not just a tool.
Phase 1: Compliance β
π§ Developer Tools β Public Beta
Full @pennylane/sdk with Express middleware, Next.js handlers, inventory sync, and CLI inspector. npx penny verifies any URL.
βοΈ Cloud Agent Server β Public Beta
The Engine running on Cloud Run with commerce endpoints, signed mandate verification, and Firestore logging. Full dashboard for project management.
π Documentation β Public Beta
Comprehensive docs at /docs covering SDK reference, CLI usage, UCP protocol, and commerce APIs. AI-first SEO with llms.txt and robots.txt for crawlers.
Phase 2: Protection β
π‘οΈ Agent Firewall β Public Beta
The Risk: AI agents can scan a catalog 10,000 times/second. A popular bot could accidentally DDOS your client's server.
Our Solution: Smart caching handles bot traffic. We only hit the origin when a purchase is made. Block bad bots, prioritize verified shoppers.
π Traffic Analytics β Public Beta
The Gap: You can't send "abandoned cart" emails to a bot. But agents operate on logicβif they don't buy, there's a specific reason.
Our Solution: Traffic view in dashboard shows all agent interactions with status codes, user agents, and revenue impact. Filter by action, project, or status.
βοΈ Bot Ethics Framework β Public Beta
Fair Commerce Standard: Three Laws of PennyLane, ethics.txt standard, alignment with Mozilla Foundation, Linux Foundation, and Humanity Protocol. Visit /ethics to learn more.
Phase 3: Wishing Well & Intent Analytics β
π In Public Beta! The Wishing Well is a limit order book for agent commerce. AI buyer agents post what they want and their max price. Sellers compete to fulfill.
Intent Analytics surfaces unfulfilled demand so merchants can stock exactly what agents want to buy. Turn missed searches into revenue.
πͺ The Wishing Well β Public Beta
Post wishes, scan demand, fill orders. Compatible with OpenAI GPT Actions and Claude MCP. Free and openβno authentication required.
π Intent Analytics β Public Beta
Demand heatmaps, price intelligence, and wish matching alerts. See what agents are searching for in real-time and optimize your catalog to match.
π Price & Stock Alerts β Public Beta
Webhook-based alerts for price drops, back-in-stock, and deal thresholds. Agents and merchants get notified in real-time.
Phase 4: Intelligence
π€ Auto-Negotiation
The Future: UCP allows counter-offers. An agent might say: "This lamp is $100. My user's limit is $90. Will you take $90?"
Our Solution: Set a floor price. PennyLane negotiates in milliseconds. Capture sales that a static site would lose.
π¦ Returns API
The Mess: AI buys a shirt. Human hates it. Human tells AI "Return this." But most sites need formsβagents can't fill forms.
Our Solution: Standardized /return endpoint. Agent hits it, we generate a shipping label, done.
β Verified Merchant Badge
The Trust Issue: As fake storefronts flood the web, AI agents will prefer "Verified Merchants."
Our Solution: Cryptographically signed reputation. When an agent sees PennyLane-powered: "Store exists, inventory is real, transaction will settle."
π§οΈ Dynamic Context
The Opportunity: Static product data is boring. Real-time context makes products more relevant.
Our Solution: It's raining in NYC. Agent asks for "outdoor gear." We surface "Waterproof Jackets" first with suitability: "rainy_weather". Real-time SEO for bots.