Let AI agents request orders.
On your terms.

Connect once; we translate ACP + Google UCP for you and enforce policy, approvals, and audit for agent order requests—before anything reaches your checkout.

Agents send an order request; you decide approve / clarify / reject before it hits checkout.

Built for grocery and retail commerce teams that need control, not hype.

AI Agent → ChatBasket (Policy + approvals layer) → Your checkout / OMS / ERP

Who this is for

For grocery and retail commerce ops teams who need explicit control over agent-originated orders.

Primary customer

Merchants - grocery and retail commerce ops teams who need to prevent operational disasters: fewer bad substitutions, fewer cancellations, fewer refunds.

Teams we work with

  • Digital commerce / e-commerce ops
  • Risk, compliance, and fraud review
  • Engineering teams integrating existing checkout and OMS
How it works

AI agents don't go through your storefront UI. ChatBasket treats every request as order intent and applies your policies and approvals before anything becomes an order.

  • No order is created until your rules are met
  • Approve, reject, or request clarification up front
  • Only approved intent reaches checkout or OMS
Agent intent → ChatBasket (Policy + approvals layer) → Merchant creates the order

View the full flow →

What stays with your systems
  • Pricing, inventory, substitutions, and fulfillment logic
  • Checkout, OMS, ERP, and payment service providers
  • Regulatory, tax, and compliance enforcement

ChatBasket is the policy and approval layer that sits before checkout, deciding whether an agent’s intent is eligible to become an order.
See the intent → decision flow →

Rules you already use

Define substitution rules, out-of-stock handling, delivery windows, age-restricted items, and approval thresholds for agent-originated orders.

Review queue for exceptions

Route high-risk or ambiguous orders for review before anything reaches your checkout—like chargeback or risk review, but for agent orders. Approve, reject, or request clarification.

Protocol-Native (ACP & UCP)

Support major agentic commerce protocols including Google's Universal Commerce Protocol (UCP) and ACP. Use standard agent-commerce semantics (intent, confirm, clarify). Add governance without forcing proprietary extensions.

Works with your checkout + OMS

Route only approved intent into your existing checkout, OMS, or ERP—and relay order status updates back to agents. You keep ownership of your pricing, your inventory truth, your fulfillment promises.

Clear reasons + audit log

Every request is logged with clear outcomes: why it was blocked, who approved it, what changed. Replay tooling helps audit and test policy changes. Make reviews and disputes traceable end-to-end.

Different rules per channel

Apply different policies per agent and channel. Example: stricter limits for new agents; looser for trusted channels. Stay flexible as new agents emerge.

Merchant Safety FAQ

Answers to the questions commerce and risk teams ask first.

Do agents place orders?

No. Agents submit order intent. Only approved intent is forwarded to your checkout or OMS.

When does it become an order?

After approval. Only after your policy passes (and a human approves when required). Until then, nothing is an order.

Do agents get order status updates?

Yes. Your systems emit order status events, and ChatBasket relays them back to the agent through ACP/UCP-style updates.

What if information is missing?

We return 'clarify'. ChatBasket returns an explicit decision with reason codes so agents can correct inputs or route for review. See the full flow →

How should we roll this out?

Start conservative, expand gradually. Start with one fulfillment method (pickup or delivery) and a conservative policy set. Expand as confidence grows.