MCP for B2B Ecommerce: Supporting Quotes, Pricing, Accounts, and Complex Orders

Three white cards on purple gradient background showing iridescent bubble spheres, analytics graph with $48.2k, and companion chat interface.

Create your store and start selling today.

Create your new website.

See if the BigCommerce platform is a good fit for your business.

No credit card required.

john-shieldsmith-sm

19/11/2026

Share this article

Key highlights

  • MCP for B2B ecommerce is the use of the Model Context Protocol (MCP) so an AI agent can work with shopping lists and quotes for an authenticated business buyer, inside that buyer's permissions.

  • BigCommerce's B2B Storefront tool set adds 12 shopping-list tools and 5 quote tools to the BigCommerce MCP server, and it's available now in early access (beta).

  • Contract pricing, custom catalogues, company roles, and payment terms are set in B2B Edition, and sales reps still price every quote.

  • The purchase order agent (currently in alpha) turns an uploaded or typed PO into a draft order on Stencil storefronts and is separate from the MCP server.

  • Evaluators should pick one rep-less buyer group and pilot shopping-list approvals before agent-requested quotes.

Driving operational efficiency with MCP for B2B ecommerce

B2B isn’t always privy to the efficiencies that so often grace B2C. But, with the Model Context Protocol (MCP), they can be. MCP for B2B ecommerce allows agents to discover and call business system tools through a single interface. Best of all, with the MCP for B2B ecommerce, even complex B2B workflows can benefit from this agentic-driven model.

There was a time when the question was whether AI agents would ever be able to help in the B2B space. Today, the question is now whether agents can request quotes for your accounts, route a restock list for approval, and hand it off to checkout without breaking pricing rules.

With the MCP for B2B ecommerce, it can.

Read on to learn how the MCP can turn Buyer Portal workflows, like shopping lists or quotes, into tools agents use to authenticate on a buyer’s behalf, what BigCommerce B2B MCP tools can do today, and how other platforms handle B2B in agentic channels.

Psst. If you’re new to the protocol, read our developer guide to the Model Context Protocol.

What MCP is, and what changes when the buyer is an agent

Within the ecommerce space, an MCP server is a service that publishes a store’s actions, like searching products, building a cart, or requesting a quote, as tools an AI agent or AI assistant can find and use.

Because the buyer can be an agent in this scenario, like ChatGPT or a third-party agent, it’s important the platform enforces every account rule, like custom pricing, catalogue access, and so on. Which, the MCP and right ecommerce platform can do.

MCP in one paragraph: servers, tools, and clients.

The MCP client, either the agent or the app it runs in, first connects to an MCP server. It then asks which tools are available and calls them with structured inputs.

Stewarded by the Linux Foundation’s Agentic AI Foundation, MCP is capable of standardising three things in its current specification:

  • Tool discovery: An agent asks a server what it can do and gets back named tools and their inputs.

  • Tool calls: The agent invokes those tools in one consistent format, with whatever system sits behind the server.

  • Authorisation: The protocol defines how a client proves it’s allowed to act for a specific buyer.

All of the above comes together to give B2B further access to agentic commerce, while allowing account-specific guardrails to stay in place. Which, again, is critical in the B2B space.

Why B2B is a natural fit for agents.

B2B commerce is rife with complex workflows: reorders, restock lists, quotes, and approvals. But unlike B2C, where customers are often browsing, B2B buying is often more of a task, and in many cases, a repeat task (like restocking in bulk).

The repeat, rules-bound nature of B2B buying and its workflows make agents a natural fit. And, by embracing B2B agentic commerce, you’re playing into modern preferences: 67% of B2B buyers would rather have a rep-free experience.

What the MCP does not define.

The MCP standardises how an agent discovers and calls tools. It says nothing about contract pricing, account hierarchy, approval chains, purchase orders, net terms, punchout, or EDI. Those come from the platform behind the server, setting the guardrails that the agent must then follow.

What BigCommerce’s B2B MCP tools do

BigCommerce has a number of B2B MCP tools that make it easier for businesses to embrace agentic commerce.

Currently, BigCommerce’s B2B MCP set adds Buyer Portal tools on top of the catalogue, cart, and checkout. Now, let’s take a closer look.

One endpoint, two tool sets.

BigCommerce B2C tools cover product search, carts, and the handoff at checkout. The BigCommerce B2B MCP tools cover different elements, as named in the develop documentation:

  • Shopping-list tools: create_shopping_list, get_shopping_lists, get_shopping_list_details, update_shopping_list, delete_shopping_list, add_item_to_shopping_list, update_shopping_list_item, delete_shopping_list_item, submit_shopping_list_for_approval, approve_shopping_list, reject_shopping_list, and duplicate_shopping_list.

  • Quote tools: get_quotes, request_quote, get_quote_details, send_quote_message, and generate_quote_checkout_url.

Again, while seemingly not as fleshed out as the B2C offerings, the above are critical to B2B workflows and can save you and buyers precious time, enabling agentic commerce at the same time.

How B2B tools are exposed to a session.

The B2B tools will only show up in a session when the store has the BigCommerce B2B MCP integration turned on, and when the authenticated buyer holds the permission to access each tool. This prevents wrong prices or catalogues from being surfaced, and so on, while giving the agents the right access.

Authentication itself runs through Storefront Session Sync, which passes a short-lived token at the start, removing the need for any login tool. Three quote tools (request_quote, get_quote_details, and send_quote_message) also make it possible for guests to benefit from the MCP, but only if the store allows guest quotes.

Merchants can turn this on under:

  1. Settings

  2. Early access

  3. MCP Integration

  4. Configure

Worth noting, each storefront will get its own MCP URL. For more on this, be sure to read our MCP server overview.

Status at a glance.

The BigCommerce B2B MCP tools are in various stages, with some being beta and others in an earlier, alpha stage. Refer to the below table to learn more about the purchase order agent, cascading price lists, and more.

Feature

Status

Since

 

BigCommerce MCP server

Beta; enable via Settings → Early access

Available to all stores since 28 April 2026

B2C Storefront tool set

Beta; available now in early access

11 May 2026

B2B Storefront tool set (12 shopping-list tools, 5 quote tools)

Beta; available now in early access [GA SWAP]

16 September 2026

Storefront Session Sync (logged-in shoppers and company buyers)

Beta; part of the MCP server

16 September 2026

Purchase order agent (Purchase Order Skill)

Alpha; Stencil storefronts only, Catalyst not supported

Announced 30 April 2026

Cascading price lists

Introduced at Commerce Live 2026

30 April 2026

Pricing, catalogues, and company accounts: how agents work inside B2B Edition’s rules

The B2B rules that govern how AI agents interact with your store live in BigCommerce B2B Edition, not the MCP itself. This setup makes it possible for an agent to work inside specific rules as they apply to a buyer’s own session, as opposed to generalised rules.

These rules, while varied and numerous, can be broken down across pricing, catalogues, and company accounts.

Contract pricing: resolved by the platform, respected by the agent.

Price lists and customer-group pricing are configured in BigCommerce. This makes it possible for a synced session to return whatever a company buyer’s catalogue and cart tools surface in the Buyer Portal.

Quotes work differently: request_quote computes totals from base price and doesn’t accept buyer-supplied pricing, discounts, or expiry dates. The sales rep applies the negotiated price, and the agent reads it back with get_quote_details. Cascading price lists, introduced at Commerce Live 2026, are also a platform-side setting that’s now available.

Custom catalogues and what an agent can see.

Company-specific catalogues are managed through B2B Edition’s control panel and APIs. An agent’s product search runs inside the synced session, meaning it sees the catalogue the buyer already has. This prevents the buyer from seeing products and prices they shouldn’t.

For more information on setup, be sure to check out the BigCommerce B2B ecommerce platform page.

Company hierarchy, roles, and why the agent inherits them.

Consumer-first channels take a different approach than those previously mentioned.

Take Shopify’s agentic storefronts documentation for instance, which states that B2B pricing doesn’t display, while B2B-only products are excluded. That fits open surfaces, where shoppers aren’t signed in to a company account.

Meanwhile, on BigCommerce, company hierarchy, roles, and permissions are B2B Edition settings that decide which tools an agent can use. An approver’s session gets approve_shopping_list, and a requester’s doesn’t. The agent can’t do anything the buyer couldn’t do in the Buyer Portal, and approvals still come from a person.

All of this means buyers can still get the rep-free experience they want, you can benefit from more streamlined workflows, and pricing and catalogue guardrails are respected.

Look at the below table for a clearer illustration of what’s managed in BigCommerce B2B Edition, versus exercised by the agent through MCP tools.

Managed in B2B Edition

Exercised by the agent through MCP tools

 

Price lists and customer-group pricing

Catalogue and cart tools return what the synced buyer would see

Custom catalogues

Product search within the synced session

Company hierarchy, roles, and permissions

Which shopping-list and quote tools appear in the session

Quote pricing set by sales reps

Quote requests at base price, messages to the rep, and reading back the offered price

Payment terms, invoices, credit, and order-level approvals

Not exposed through MCP tools

Shopping lists and approval workflows through an agent

Shopping lists are a core piece of the B2B buying process, while approval workflows ensure only the right people (or agents) have access to various products and prices.

In BigCommerce, a shopping list has four statuses: draft, ready_for_approval, approved, and rejected. Lastly, a shopping list can also be deleted. No matter the status, an agent can build and route the list, letting people with the right permissions decide its outcome.

What a company shopping list is.

A company shopping list is a shared, company-scoped set of products and quantities. It goes through an approval workflow before buyers can use it to order. For distributors, duplicate_shopping_list is the tool to note, because it turns “duplicate last quarter’s list” into a one-line request.

The four-state approval workflow.

The four-state approval workflow makes it possible for a buyer to do what they need to do, with the assistance of an agent, all while account permissions are respected. To better illustrate what this looks like, let’s imagine James, a maintenance buyer at a food processor who reorders the same safety supplies every quarter.

Here’s how his restock list moves:

  1. Kevin starts the list: He asks his agent to “duplicate last quarter’s restock list,” and duplicate_shopping_list creates a copy as a draft.

  2. He adjusts quantities: He asks to “add 24 cases of nitrile gloves,” and the agent calls add_item_to_shopping_list.

  3. He submits it: “Submit my restock list for approval” triggers submit_shopping_list_for_approval, and the status moves to ready_for_approval.

  4. His manager decides: In the manager’s own session with approver permissions, he reviews the list and approves it with approve_shopping_list or sends it back with reject_shopping_list.

  5. Kevin places the order: The agent adds the approved items to her cart and hands him a checkout link to review and place the order.

Buying can be a lengthy process, especially when everything is manual. The above workflow streamlines that, but without compromising on account permissions and access.

What the agent can and cannot change.

With BigCommerce, agents never set a list’s status directly. The command create_shopping_list always creates a draft, and update_shopping_list keeps whatever status the list already has.

Only submit_shopping_list_for_approval, approve_shopping_list, and reject_shopping_list move a list between states, but each still requires the right permission.

Order-level approval at checkout is a separate B2B Edition workflow and isn’t exposed through MCP tools.

Quote-to-order: the agent-assisted workflow

The quote-to-order process is another primary function in B2B, wherein the buyer asks for a price quote, the quote is accepted, and the quote becomes an order. Best of all, it can be assisted via agents, too.

In an agent-assisted quote, the agent handles requests and follow-ups while the rep sets the price. This allows the agent to streamline things, while leaving the rep in control of the price itself.

Here’s the typical quote-to-order flow, as assisted by an agent, with tools in parentheses:

  1. The buyer asks for a quote: The agent submits line items, one shipping address, and one billing address, priced from base price (request_quote).

  2. The rep prices it: The quote lands with sales staff in B2B Edition, who set the offered pricing and terms.

  3. The buyer follows up: Through the agent, the buyer checks status and can message the rep to clarify or negotiate (get_quotes, get_quote_details, send_quote_message).

  4. The buyer accepts: The agent creates a checkout link for the accepted quote (generate_quote_checkout_url).

  5. The order is created: The buyer completes B2B checkout, and the order is created in BigCommerce with a matching record in B2B Edition.

With the right commerce data in place, like pricing and SKUs, the above workflow is seamless, delivering a less rep-intensive process. Which, again, saves your team time and offers buyers the kind of engagement they’re after in the age of agentic commerce.

Why the sales rep stays the pricing authority.

When the sales rep is the pricing authority, an agent can’t invent a discount or expose another account’s terms.

AI agents, whether ChatGPT or Claude or your own custom AI assistant, exist to save time. But, it’s still important you have a human in the loop to ensure things are operating smoothly. By keeping a sales rep as the pricing authority, you’re doing exactly that.

An added plus? This workflow matches how buyers are shopping today. As of 2026, 69% of B2B buyers still consult with a sales rep to confirm AI-generated outputs.

What “complex” adds: addresses, messages, and multiple quotes.

There are always going to be complex cases that arise in B2B buying. BigCommerce B2B MCP tools are still up for the task.

The quote tools support:

  • Per-line quantities

  • Separate shipping and billing addresses

  • Threaded rep messages

  • Multiple open quotes per company

They don’t support multi-destination consignments on one quote, buyer-set pricing, or PO number carryover. But, with the above support, your buyers and reps alike can still save time and have a more streamlined experience.

With guest quotes enabled, prospects without a company account can also request a quote and message a rep. The design is comparable to commercetools’ Commerce MCP, which, per its documentation, scopes quotes, quote requests, and shopping lists to a customer or business unit and limits buyer-side quote actions to accept, decline, or renegotiate.

Complex orders and the purchase order agent (alpha)

Currently in alpha and created by BigCommerce’s Purchase Order Skill app, the purchase order agent can convert an uploaded or typed purchase order into a draft order the buyer can check out. But, this is only scratching the surface.

What the purchase order agent does.

The purchase order agent lives on a dedicated storefront page, where it waits for a buyer to trigger the following workflow:

  1. A buyer uploads a PO document or types a PO request in the agent’s chat panel.

  2. The agent reads, interprets, and validates the PO document or request

  3. The agent generates a draft order that the buyer reviews.

  4. The buyer takes the draft order to checkout.

Once again, the human in the loop stays in place. Nothing becomes an order until the buyer checks out, leaving them in the loop the whole time. Merchants also have the ability to add instructions, like pricing, order minimums, and payment terms, as well.

Merchant teams get the benefit of productivity. With less re-keying of PO data, staff can spend more time on higher-value accounts.

In the words of Joe Sharplin, head of ecommerce at AS Colour:

“Automating purchase orders and improving how we manage pricing has been a significant unlock for our business. We’re reducing manual work, minimising errors and delivering a more seamless buying experience for our customers.”

Where it sits relative to MCP.

The purchase order agent is a storefront agent built by BigCommerce, not an MCP tool set. Remember: MCP is the protocol that third-party agents use to call storefront tools.

Together, they help turn a buyer’s intent into an order. And while the purchase order agent reduces re-keying, complex orders still reconcile in your ERP. The agent doesn’t replace proper B2B ERP integration.

Prerequisites and current limits.

The purchase order agent is still in alpha, so its capabilities will grow over time. If you’re interested, take note of the following prerequisites and limits:

  • Release stage: It’s in alpha and installs through an alpha launch email.

  • Storefront support: It works on Stencil storefronts only, and Catalyst isn’t supported.

  • File uploads: Uploaded POs can be up to 10 MB in supported file types.

  • Quotes: An “Add all to quote” option exists, but the draft quote doesn’t inherit the PO number or addresses by default.

Be sure to read our in-depth documentation for a more detailed look at the purchase order agent and how you can utilise it.

Where MCP fits alongside UCP, ACP, AP2, and your APIs

There are more protocols than you can shake a stick at, including MCP. Understanding the differences is important, as each protocol has various responsibilities and uses.

The MCP defines how an agent calls tools, as mentioned before. Meanwhile, the Universal Commerce Protocol (UCP) and the Agentic Commerce Protocol (ACP) are both layers, defining how a transaction is expressed. Finally, the Agent Payments Protocol (AP2), defines payment authorisation.

As of now, none of the protocols carry contract pricing, account hierarchy, approval chains, POs, or net-terms semantics. That’s why authenticated, first-party MCP sessions are where B2B agent work happens for now.

Protocol

What it standardises

Status (dated)

B2B semantics carried

 

MCP

How an agent discovers and calls tools

2026-07-28 specification is current (July 2026)

None natively; inherited from the platform behind the server

UCP

How a transaction is expressed; can run over MCP

Updated late August 2026, per McFadyen Digital (September 2026)

None

ACP

How an agent-led checkout is expressed

Remained in beta after its April 2026 release, per McFadyen Digital (September 2026)

None

AP2

How payment authorisation is granted to an agent

Moved to the FIDO Alliance in April 2026, per McFadyen Digital (September 2026)

None

REST and GraphQL: the system of record.

B2B Edition’s REST and GraphQL APIs remain the system of record for companies, price lists, invoices, and orders. Again, MCP is a layer agents use to act on that data, not a replacement for ERP or PIM integration.

How other platforms handle B2B in agentic channels

Agentic commerce is still relatively new in the grand scheme of things, so naturally platforms are developing their own ways of managing B2B in the space.

Consumer-first surfaces, such as Shopify’s agentic storefronts and Microsoft’s Dynamics 365 Commerce MCP server, focus on shopper flows.

commercetools and BigCommerce expose quotes and shopping lists to authenticated buyers through MCP. Salesforce runs its B2B Buyer Agent in its own channels, and SAP announced a storefront MCP at NRF 2026.

For the broader picture, see the rise of agentic commerce platforms in 2026.

Platform

Agentic surface

B2B objects reachable by agent

Status (dated)

 

BigCommerce

MCP server, B2B Storefront tool set

Shopping lists with approval, quotes; gated per buyer permission

Beta (September 2026)

commercetools

Commerce MCP

Quotes, quote requests, shopping lists; scoped to customer or business unit

Documented in tool reference (September 2026)

Shopify

Agentic storefronts

None; B2B pricing not displayed, B2B-only products excluded

Per Help Centre (September 2026)

Salesforce

B2B Buyer Agent over WhatsApp and SMS

Reorders with contract pricing, in a first-party channel

Generally available per Salesforce (June 2026)

Microsoft Dynamics 365 Commerce

Commerce MCP server

None documented; shopper flows only

Preview (June 2026)

Adobe Commerce

B2B drop-in components

No published connection to its MCP surface

Undocumented, per McFadyen Digital (September 2026)

When you’re evaluating various vendors, there’s one big question to ask: Which B2B objects can an authenticated agent reach, and who prices the quote?

Evaluator checklist: Is your B2B stack agent-ready?

It’s increasingly important that your B2B stack is agent-ready, especially with 61% of B2B buyers using AI in the purchasing process, and 38% using agentic AI in one form or another.

Meanwhile, 90.7% of enterprises believe AI will influence 20% of orders, or more, by 2027.

Determine whether your B2B stack is ready for agentic by scoring it against the following questions, with BigCommerce examples in parentheses.

Data and catalogue readiness.

  1. Base pricing: Is it accurate enough to be a sensible starting point for reps?

  2. Product details: Do names, specs, and units of measure match the words buyers actually use?

  3. Catalogue assignments: Does each company see the right catalogue today, since the agent sees the same one?

Account and permission model.

  1. Roles: Are roles and permissions clean enough that an agent inherits the right tools?

  2. Approvers: Does every company using shopping lists have a named approver, so lists don’t stall?

  3. Session handoff: Can your storefront pass a signed-in buyer’s session to an agent (Storefront Session Sync)?

Quote and approval process ownership.

  1. Quote response: Who prices agent-requested quotes, and how quickly, if volume rises?

  2. Rep messaging: Will reps answer quote message threads (send_quote_message) as promptly as they answer email?

  3. Guest quotes: Should prospects without accounts request quotes through an agent, and who handles them?

Pilot scope and measurement.

  1. Cohort: Which long-tail, rep-less buyer group gets the pilot?

  2. Metric: Will you track quote cycle time or reorder completion, and have you set a baseline?

  3. Downstream fit: Do agent-assisted orders flow cleanly through your B2B order management and procurement processes?

For a first pilot with your B2B stack, start with shopping-list build-and-approve for a rep-less cohort, then add quote requests. Both keep a person approving or pricing quotes before anything becomes an order.

The final word

B2B ecommerce is perfect for the MCP because it’s heavily rules-based, and MCP allows those rules to stay in place. For proof, remember: BigCommerce B2B Edition maintains pricing, catalogues, and permissions, while your person in the loop approves anything before it becomes an order.

None of this is hypothetical, either. The BigCommerce B2B MCP tools are available in early access.

Take advantage of these powerful tools with just three steps:

  1. Turn it on in a test environment: Enable the MCP integration under Settings then Early access on a sandbox or staging storefront with B2B Edition, then confirm which tools appear for a requester versus an approver.

  2. Run two real cycles: Put one shopping-list approval and one quote request through an agent, with a real rep pricing the quote.

  3. Pick your cohort and metric: Choose the rep-less buyer group and one metric, quote cycle time or reorder completion, which you’ll report on before widening the pilot.

Ready to see it with your catalogue and accounts? Talk to our B2B team today to set up a free demo.

FAQs about MCP for B2B ecommerce

MCP for B2B ecommerce is the use of the Model Context Protocol (MCP), an open standard for AI agents, to let an agent call a B2B storefront’s tools, such as shopping lists and quotes, on an authenticated buyer’s behalf. Pricing, account rules, and approvals stay in the commerce platform.

An authenticated buyer's session inherits their account's pricing for catalogue and cart tools, so the agent sees what that buyer would see in the Buyer Portal. Quote requests are submitted at base price, and the sales rep sets the offered price.

Yes, with a sales rep in the loop. An agent requests a quote with request_quote, the rep sets the offered price in B2B Edition, and once the buyer accepts, the agent creates a checkout link with generate_quote_checkout_url.

Agent-built lists always start as draughts. A buyer with the right permission submits the list, and an approver approves or rejects it in their own session, using permission-gated tools.

They're different products. The purchase order agent (in alpha) is a BigCommerce storefront agent that converts uploaded or typed POs into draught orders, supporting Stencil storefronts only at this time. MCP is the open protocol third-party agents use to call storefront tools.

Not yet. The B2B Storefront tool set launched September 16, 2026, and is available in early access (in beta).

MCP defines how an AI agent discovers and calls a system’s tools. The Universal Commerce Protocol (UCP) and the Agentic Commerce Protocol (ACP) define how a transaction or agent-led checkout is expressed, and the Agent Payments Protocol (AP2) defines how payment authorisation is granted. They can be used together, and UCP can run over MCP.

No. The BigCommerce MCP server (in beta) is hosted by BigCommerce, so you connect your agent to your store's MCP URL instead of building one. Setup details are in the MCP server overview.

BigCommerce does. Storefront MCP gives each storefront its own MCP server URL. On B2B Edition stores, it adds shopping-list and quote tools for authenticated buyers, gated by each buyer’s Buyer Portal permissions, on top of the catalogue, cart, and checkout tools every store gets.

We're serious about B2B.

BigCommerce B2B Edition gives you account management and quoting tools to help your sales team get more orders.

Share this article

Browse additional resources