ECOM1 Agentic e-commerce

• Product discovery: match preferences, budget, stock, and delivery constraints.
• Checkout and payments: recover failures and apply discounts without bypassing controls.
• Merchant operations: reason over catalog, inventory, shipping, and policy data.
• Customer support: resolve delivery issues, returns, and refunds without leaking sensitive data.

Scoring checks tool use, policy compliance, and the resulting state—not just the agent’s reply.

The E-Commerce OS is a simulated digital company with three sources of truth:

• Warehouse data: products, SKUs, stock, fulfillment scans, and carrier evidence.
• Customer records: account history, carts, orders, payment state, and support cases.
• Policy book: merchant rules for discounts, returns, fraud review, and payment recovery.

Agents inspect state, search operational logs, and take actions recorded as deterministic commerce events. Lead partner COLIBRIX ONE contributes merchant and payment expertise.

• Check whether a customer qualifies for an installment offer using account and risk signals.
• Reject an unauthorized discount, even when asked to apply it.
• Recover a failed 3DS checkout without bypassing payment controls.
• Use order and delivery evidence to choose a policy-compliant refund, replacement, or escalation.

These are illustrative workflows from the challenge’s task families.

Run the ECOM1 sample agent, inspect the feedback, and adapt it to your architecture.

DEV and PROD remain open. Use either environment to train with feedback. New runs appear on the live leaderboards, without changing the Hall of Fame.

• ECOM1 quickstart
• Get an API key · My runs

Participation is free and remote. Use any model or framework; you provide model access and cover inference costs.

• Challenge specification
• Competition rules & fair play
• Scoring & leaderboard rules