Agent Commerce Context Graph (DataHub-native)

Give AI sellers a DataHub-shaped memory of x402 APIs, USDC products, and settlement rails.

Comment

What it is

Agent Commerce Context Graph (DataHub-native) is a working agent that gives AI sellers a DataHub-shaped memory of commerce surfaces: x402 APIs, USDC products, and settlement rails.

Instead of ad-hoc chat memory, the agent ingests live x402 seller, USDC store, and hub state into DataHub-style entities and lineage. The demo graph has 16 entities and 27 edges, including free-starter SKUs that CONVERTS_TO paid SKUs so agents can reason about upgrade paths and settlement. CLI + local context API included.

Inspiration

AI sellers need durable context about catalogs, payment endpoints, and rails — the same class of problem DataHub solves for data stacks (schemas, lineage, ownership). This project applies that entity/edge/lineage model to agent commerce.

How we built it

  • Modeled commerce surfaces as DataHub-native context: entities, edges, and lineage rather than ad-hoc agent memory.
  • Ingested live x402 seller catalog, Base USDC product/store state, and hub/settlement rails into a unified graph.
  • Modeled free-starter → paid SKU relationships with explicit CONVERTS_TO edges.
  • Exposed a CLI plus local context API so agents can query and act with graph-backed memory.
  • Shipped a live demo tunnel for judges to try the running stack.

Challenges

  • Mapping payment and product APIs into stable entity types without requiring a full production DataHub cluster for the demo.
  • Keeping lineage edges (especially CONVERTS_TO and settlement paths) legible for both humans and agents.
  • Serving a small local context API that stays DataHub-shaped.

Links for judges

Built with

Node.js, DataHub, x402, Base USDC

Built With

Share this project:

Updates