Skip to content
AI/MCPE-commerce SaaS

Built an MCP server and an embedded AI agent that let store owners run their Sendd stores — inventory, orders, customers — by chatting, including creating many listings in one go.

Client
Sendd
Role
Co-developer — MCP server & AI agent
Year
2025–present
Stack
Ruby on Rails, Inertia.js, React, Model Context Protocol, PostgreSQL, Redis, Google Cloud
Making Sendd agent-native: an MCP server and AI agent for commerce — cover (placeholder)

Context & problem

Sendd lets a business launch a complete e-commerce store in about 60 seconds — selling physical products, digital products, event tickets, subscriptions, enquiries, rentals, food and beverage, affiliate links, or a custom-price option for in-person payments and donations. Its companion product, Sendd Market, lets store owners list those products on a marketplace and get paid their share while the marketplace owner earns a commission.

Launching is fast, but day-to-day management wasn't. Store owners wanted to create and edit listings quickly, migrate catalogs in from platforms like Shopify, and run the store without clicking through admin screens for every change. The opportunity was to let them manage the whole store the way they already work — by chatting — and to let their own AI tools connect to Sendd directly.

Approach

An MCP server over the whole store

We exposed Sendd's store operations through a Model Context Protocol server, so any MCP-compatible client — or Sendd's own agent — can drive a store with full CRUD over inventory, orders, and customers. Each capability is a typed tool with an explicit schema, scoped to the authenticated merchant's store.

Because the surface speaks MCP, store owners aren't locked into our agent: they can point their own assistant at their store and it just works.

typescript
// One call, many listings — the schema accepts a batch
{
  name: "create_listings",
  description:
    "Create one or more product listings on the authenticated merchant's store.",
  inputSchema: {
    type: "object",
    properties: {
      listings: {
        type: "array",
        description:
          "Each listing: physical, digital, ticket, subscription, " +
          "enquiry, rental, F&B, affiliate, or custom-price.",
        items: { $ref: "#/$defs/listing" },
      },
    },
    required: ["listings"],
  },
}

An agent built into the product

On top of that surface we embedded an AI agent directly in Sendd. Owners describe what they want in plain language and the agent does it — whether that's importing a catalog from another platform, editing inventory, or handling orders and customers.

The feature I'm proudest of is bulk listing creation: an owner can spin up many listings in a single conversation instead of filling out one form at a time, which turns a tedious migration into a few sentences.

Bring your own model

The agent is billed by usage, so there's no per-seat or monthly add-on. And if an owner would rather not use our models or pricing, they can plug in their own model of choice — they keep full control over cost and capability without giving up the integration.

Outcome

Sendd is live and global at sendd.store and sendd.market — anyone can start selling, and store owners can now manage their store conversationally instead of screen by screen.

The economics are the other half of the story. Sendd charges a flat 1% of sales, with no monthly subscription and no hosting fee. For comparison, a Shopify Basic plan is $39/month (around $468/year) before payment-processing and app costs, and a production WooCommerce store typically runs $1,800–$15,000+ per year once hosting, extensions, and maintenance are added. For a store owner, that recurring overhead effectively disappears.

60sTo launch a full store
1%Of sales — no monthly or hosting fee
9Product types sellable out of the box
[OWNER: paste a real quote — what they hired me for and the result.]
[OWNER: Name][OWNER: Role, Company]
Sendd AI agent (placeholder)
Sendd Market storefront (placeholder)

Have a project that needs a senior pair of hands?

Tell me what you're building. I'll tell you honestly whether I'm the right fit and how I'd approach it.