BuyingMesh starts with a practical observation: the world's supply is not missing. It is difficult for software to understand.
Factories, trading companies and wholesale markets already contain the products that global buyers need. But their information is spread across storefronts, spreadsheets, chat messages and private relationships. A human can spend hours translating that information into a purchase decision. An AI agent needs something different: stable product identities, comparable commercial fields, supplier context, explicit approval boundaries and machine-readable settlement states.
Our ambition is to build that missing layer.
Start with Yiwu, build for the region
Yiwu is the starting point because it compresses a remarkable amount of supply into one commercial ecosystem. It gives us a useful environment to learn how product facts change, how merchants confirm availability, how variants should be represented and how an order moves from an agent's request to a supplier's operation.
The goal is not to create another consumer marketplace. BuyingMesh is building infrastructure for AI agents and B2B buying teams. A supplier's catalog should be available as structured data, and a buyer should be able to ask an agent for a decision grounded in SKU, material, dimensions, MOQ, price, inventory, lead time, qualification and freshness.
Once the data and operating model work in one dense sourcing market, the network can expand across East Asia and Southeast Asia. Each new region will add local supplier and logistics knowledge while keeping the same agent-readable contract.
The data layer for agent commerce
Today, an agent often has to scrape a page, infer whether a price is current, guess which product variant is being described and ask a human to reconcile conflicting information. This is fragile and difficult to audit.
BuyingMesh turns a supplier's catalog into a normalized record with a stable SKU identity and an explicit evidence trail. The record can include product identity, GTIN and variant attributes; material, dimensions and packaging; MOQ, quantity breaks, price and currency; current inventory and expected lead time; supplier qualification and review status; source reference, confidence and last-confirmed time; and product images that a human buyer can use for final confirmation.
The same record is available through REST and MCP. This gives an agent a predictable way to discover supply, compare offers and explain its recommendation. It also gives a supplier a clear standard for the data it needs to maintain.
The settlement layer for agent commerce
Discovery alone does not create a new procurement network. An agent also needs a safe way to move from a recommendation to a completed transaction.
Query → buyer approval → compliant settlement and fulfillment
The planned BuyingMesh settlement architecture separates three moments. First, an agent can pay a small fee for a structured product or supplier lookup when paid access is enabled. Second, the agent presents price, quantity, lead time, supplier identity and assumptions to the buyer, who explicitly confirms the purchase. Third, the buyer's payment is held under a compliant settlement arrangement, the supplier fulfills the order, and funds are released according to the agreed delivery and acceptance rules.
For China-related settlement, our current direction is to work through a properly licensed Hong Kong stablecoin custody and payment structure, or regulated partners that provide the same function. A buyer may pay an approved stablecoin, such as USDC, through the agent workflow. After the buyer confirms the goods and the transaction satisfies the applicable release conditions, the settlement provider would convert the funds and pay the Chinese supplier in RMB through permitted channels.
Planned architecture: This is not a currently available custodial service. Before any customer funds are accepted, the operating entity and its partners must satisfy the relevant requirements for virtual assets, custody, money services, AML/KYC, sanctions screening, foreign exchange and cross-border payment. The first production version can use regulated custody and payment partners while BuyingMesh focuses on data, approval and fulfillment interfaces.
A business model aligned with useful actions
Agent query fees
Simple discovery can remain free during the early network phase. As the catalog becomes fresher and more valuable, agents can pay a small per-query or per-result fee for structured product data, supplier qualification, real-time availability and comparison services. x402 provides a natural protocol for machine-to-machine micropayments, with the price and currency declared by the API rather than hidden in a subscription.
Transaction fees
When an order is completed, BuyingMesh can charge the seller a small marketplace infrastructure fee. A planned range of three to five percent gives the platform room to provide catalog governance, buyer confirmation, supplier notifications, payment coordination, dispute evidence and fulfillment events. The exact rate can vary by category, order value and the amount of operational support required.
Merchant data services
Suppliers that want higher visibility and better conversion can pay for data onboarding, catalog normalization, image and variant management, freshness monitoring and qualification workflows. This turns data governance into a service with a clear return: more of the supplier's inventory can be discovered and evaluated by agents.
Verified regional data
Some buying agents will need a continuously refreshed view of a category or sourcing region. BuyingMesh can offer verified data feeds, category monitoring and private availability signals under a commercial data agreement, while respecting supplier authorization and access boundaries.
Trust is the product
The network will only work if an agent can tell the difference between a product description, a current offer and a verified fulfillment commitment.
BuyingMesh keeps several distinctions visible: sandbox records are disposable test fixtures, not commercial offers; merchant listings are private until reviewed and approved; price, inventory and lead time are snapshots that require confirmation; supplier claims are separate from platform-observed performance; query payment is separate from order settlement; and buyer confirmation is required before an order mutation.
A staged path to the network
The first phase is narrow: onboard a small group of merchants, publish high-quality SKU records, validate search and comparison, and run the full sandbox flow from query to order event. The second phase introduces approved production listings, supplier-specific order notifications and confirmed fulfillment signals. The third phase adds paid queries, regulated settlement partners and more categories across the region.
We are not trying to digitize every supplier at once. We are trying to prove that one reliable agent-to-supplier path can be repeated. Each successful transaction improves the data contract, the supplier workflow and the evidence available to the next buyer.
The long-term view
BuyingMesh aims to become the data layer and settlement coordination layer for agent commerce: a network where an agent can understand what is available, compare it honestly, obtain buyer approval, pay through a compliant route and keep the supplier and buyer synchronized until delivery.
The starting point is Yiwu. The direction is East Asia and Southeast Asia. The larger change is from relationship-bound, page-oriented B2B buying to a protocol where supply, trust and fulfillment can be consumed by software without removing human judgment from the moments that matter.
Current status
The public BuyingMesh catalog and order flow are currently sandbox infrastructure. Merchant onboarding, production catalog approval, paid queries and settlement are being developed in stages. See the technical architecture article, merchant onboarding page and agent guide for the current interfaces.