Agentic Commerce

In development: an agent-led shopping and merchant-operations research system, from product discovery to an explicitly approved checkout. Announcement and demonstrations will follow as the build progresses.

Shopify UCP, MCP, A2A, TypeScript, Vercel AI SDK, Cloudflare Workers, Durable Objects, Workflows, PostgreSQL, pgvector, OpenTelemetry

In development — announcement to follow. This is an active build and research direction, not a released commerce platform.

Commerce that agents can understand

The planned system connects product discovery, product comparison, merchant tools and checkout preparation. A person explicitly approves the exact purchase before an order is submitted.

Planned technologies

Commerce protocols: Shopify Universal Commerce Protocol (UCP), Model Context Protocol (MCP) and Agent2Agent (A2A) for scoped discovery and tool handoffs.
Agent engineering: TypeScript, Vercel AI SDK, structured outputs, tool policies and evaluations.
Runtime: Cloudflare Workers, Durable Objects and Workflows for sessions, resumable jobs and idempotent orchestration.
Data: PostgreSQL, pgvector and source-linked catalog records for retrieval and comparisons.
Controls: short-lived exact-payload approvals, hosted payment handoff, audit trails, retries and OpenTelemetry tracing.

What is being worked on

Discovery quality, reliable merchant integrations, clear permission boundaries and evaluation against real product data. Providers and protocols will be confirmed through implementation; this list describes the intended architecture.

Discuss a commerce use case or start a project brief.