/api/v1 Products and images
Create, edit and delete products by code, with their variants and categories. Upload and remove photos.
GET · POST productsPUT · DELETE products/{code}POST products/{code}/images AI AGENTS · OPEN API
ChatGPT, Claude or Gemini running your wholesale store
Connect your AI assistant to the API and ask it, in your own words, for what you do by hand in the panel today: load products, raise a price list, review the day's orders, block a late payer. Everything the API allows, your AI does for you.
Works with any AI that speaks REST or MCP · Bearer token · 20 minutes to the first connection How it works
Your AI already knows how to talk to an API. It only lacked the keys.
The store's Open API exposes what you see in the panel. An AI agent reads that reference like a manual, turns your request into the right calls and shows you the result. You bring the voice; the AI, the hands.
1
You ask in your language
"Raise the Wholesale B list 8% in Cleaning and show me the table first." No spreadsheets, no menus, no exports.
2
The AI calls the API
It reads the category's 142 products, builds the "before → after" table, waits for your OK and writes the new prices per list and variant.
3
The store changes instantly
Your customers already see the new price. The AI leaves you the summary: what changed, how many records, what's pending.
Not to be confused with the AI Salesperson, which serves your customers inside the store, or with the panel's support assistant. This is your own AI, the one you already use every day, running the business from the outside through the API.
Everything the API allows
What your AI can manage in the store
Each card is an API resource. Below it, an example of what you'd ask the AI.
Prices per list
Read and write the price of every product and variant on each wholesale list, without touching the others.
GET · PUT products/{code}/pricesGET · POST price-listsOrders
List and read every order with items, totals, payment and shipping. Void and leave internal notes. Orders are born in the cart: the AI doesn't make them up.
GET ordersGET · PUT orders/{id}Customers
Onboard buyers, edit their data and tax status, assign them a list and block them. They are never deleted: the history stays.
GET · POST customersGET · PUT customers/{id}Categories
Create, rename, nest, sort and hide categories. The catalog tree, tidied by an AI that understands it.
GET · POST categoriesPUT · DELETE categories/{id}Reports and routines
Cross orders, customers and catalog to build the report you do by hand today, and leave it running on its own every week.
GET orders · customers · productsFull reference of endpoints, parameters and limits on the API page and in the help center.
With any AI
The AI you already use, through the path that suits you
They all reach the same API. The only thing that changes is the plug.
A terminal agent
Runs on your computer and calls the API directly. You leave it the token in a variable, hand it the reference and ask. No coding at all.
- Claude Code
- Codex CLI
- Gemini CLI
- GitHub Copilot
- Cursor
- Windsurf
An MCP connector
To operate from Claude, ChatGPT, Gemini or Copilot: an MCP bridge exposes the endpoints as tools. The AI itself writes it in minutes from the reference.
- Claude
- ChatGPT
- Gemini
- Copilot Studio
- Le Chat
- VS Code
A GPT or agent with actions
A "Store operator" anyone on your team uses from ChatGPT or Microsoft 365 Copilot, with an OpenAPI schema and the token as authentication.
- GPT personalizado
- Copilot Studio
- Microsoft 365 Copilot
Automations and SDKs
For what has to run on its own: a trigger, an HTTP node to the API and an AI node that summarizes or decides. Or the SDK of the provider you prefer, open models included.
- n8n
- Make
- Zapier
- Power Automate
- Anthropic · OpenAI · Google · Mistral
- Llama · DeepSeek · Qwen
A day with the AI in the store
What used to be an hour in the panel is now one sentence
Which orders came in overnight, from whom and for how much? Flag the ones without payment.
The AI lists the orders, cross-checks payment status and returns the table with totals per customer.
GET ordersGET customers/{id}The supplier's new list arrived. Update costs and raise the Wholesale A list 6% on those 80 codes.
The AI reads the spreadsheet, builds the before → after table, waits for your OK and writes the prices.
GET products/{code}PUT products/{code}/pricesLoad the 300 new products from the spreadsheet, with photos. If the code exists, update it.
The AI splits new from existing, tells you how many of each and loads them in batches within the API limit.
POST productsPOST products/{code}/imagesThese 12 customers have overdue debt. Block them and leave the reason and date on the record.
The AI finds each customer by email, confirms the list with you and blocks them one by one.
GET customersPUT customers/{id}Order 8842 is picked up Thursday and 8843 ships with the usual carrier. Note it down.
The AI writes the internal notes on each order, signed with who requested it.
PUT orders/{id}Weekly summary: orders per customer, average ticket, out-of-stock products and customers who haven't bought in 60 days.
The automation queries the API, the AI drafts the report and it reaches you on WhatsApp before you open the panel.
GET ordersGET productsGET customers How to run it
With the keys to the store, but with clear rules
An AI connected to the API has the same power as an admin. Six rules turn it into the tidiest employee you ever had.
- Read firstThe first week, only queries and reports. Writes come when you trust how it understands your store.
- Confirm before writingEvery change is announced and waits for your OK. On big batches, a preview table with before and after.
- The token never travels through the chatIt lives in an environment variable, in the MCP bridge or in the GPT's settings. It's the business's master key.
- Respect the limits120 calls per minute, pages of 100, bodies up to 5 MB. The AI paginates, waits and retries on its own.
- What the API doesn't allow, doesn't happenOrders aren't created via API and customers aren't deleted: two design limits that cap any mistake.
- Close with a summaryWhat changed, how many records, what's pending. On orders, internal notes leave the trail.
What you need
Three things, and the first connection takes 20 minutes
You pick the AI. The rest is in your panel or we activate it for you.
- The API active on your account. It's an add-on service; today it's available for businesses in Argentina and we activate it through support. If you're in another country, ask us.
- A Bearer token, generated in Settings → ERP → API Token. It's shown once and you store it like a master password.
- The AI you already use: Claude, ChatGPT, Gemini, Copilot, Cursor or an automation. The guide has the step by step for each one.
FAQ
What everyone asks before connecting their AI
Which AI does it work with?+
With any AI that can call a REST API or connect to an MCP server: Claude and Claude Code, ChatGPT and Codex, Gemini and Gemini CLI, Microsoft Copilot and GitHub Copilot, Grok, Mistral, DeepSeek, open models like Llama or Qwen, editors like Cursor and platforms like n8n, Make or Zapier. The API speaks JSON over HTTPS with a Bearer token, the format they all understand.
Do I need to know how to code?+
No. With a terminal agent (Claude Code, Codex, Gemini CLI, Copilot) it's enough to give it the token as an environment variable and the link to the reference. If a path needs code, like the MCP bridge or the OpenAPI schema, the AI itself writes it.
What can and can't it do?+
It can do everything the API allows: products and images, prices per list, categories, price lists, customers (onboarding, editing, blocking) and orders (reading, voiding, internal notes). It doesn't create orders, because they are born in the cart or the panel where prices and discounts are computed, and it doesn't delete customers: it blocks them.
Is it safe to give it access to the store?+
As safe as you are with the token: it lives outside the chat, in an environment variable or in the connector, and the AI works under rules (read first, confirm before writing, summary at close). Each token has its own rate limit, orders aren't created via API and customers aren't deleted. If you think the token leaked, support generates a new one for you.
Is there an official MCP connector?+
Not yet. Today the path is the REST API: terminal agents use it directly, and for chat apps you run an MCP bridge the AI itself writes in minutes from the reference. The guide explains how.
How much does it cost?+
The API is an add-on to your VentasxMayor plan; you pay for the AI to its provider as you already do. There's no per-call or per-volume API charge: the limit is rate-based, 120 calls per minute per token.
Start today
Hand the keys to your store to the AI you already use
We activate the API, review your first connection and leave the rules ready for your team.