For AI Agents

Let AI agents look up machines and price them directly

MAQ exposes two machine-readable endpoints: a standard REST API, and a Model Context Protocol (MCP) server. An agent can pick a machine from a single sentence of requirements, look up machines currently for sale, retrieve every configurable field and its options, submit a configuration to get an exact total price, and generate a quote URL that can be handed to the customer. Public, no sign-up, no API key.

Neither endpoint creates an order. Agents handle configuration and pricing; placing an order is always confirmed and submitted by the customer on the order page. MAQ does not accept orders placed on a customer's behalf by an agent, and never collects any payment information through the API.

Pricing conventions

All amounts are the total price in New Taiwan Dollars, tax and shipping included — no separate tax or shipping charges.
Quotes are calculated by the same configurator engine used on the order page, so the numbers match what's shown on the website.
Prices may change with availability; a quote is valid for two days — request a new one after it expires.
Only the total price for the complete machine is provided; individual component prices or option surcharges are not disclosed.

REST API

Base URL https://maq.tw/api/v1　·　60 requests per minute per IP　·　no API key, no sign-up　·　HEAD is treated the same as GET

`GET\|POST /recommend`	Pick a machine from a single sentence describing the need (use this when the customer can't state specs)
`GET /machines`	Machines currently for sale and current pricing
`GET /machines/`	Specs, plus the available options for every configurable field
`GET\|POST /quote`	Submit a configuration, get back the exact total price and configuration details
`POST /quote/create`	Save a configuration as a quote, returning a short URL that can be handed to the customer

Example: the customer only states a need, let MAQ recommend a configuration

curl -s https://maq.tw/api/v1/recommend \
  -H 'Content-Type: application/json' \
  -d '{"need":"我只是 RAG／ERP 需求","budget":300000}'

The response returns the recommended machine, its use case, the models it can run, the total price with tax and shipping, and an order_url — that link pre-fills the order page with this configuration, so the customer sees exactly the machine the agent configured.

Example: swap the Llama 70B machine's GPU for the RTX PRO 6000 96GB, then get a quote

curl -s https://maq.tw/api/v1/quote \
  -H 'Content-Type: application/json' \
  -d '{"slug":"ai-high","selection":{"graphic":"GPU_RTX_PRO_6000_96GB"}}'

Field codes (e.g. graphic) and part numbers (e.g. GPU_RTX_PRO_6000_96GB) all come from the GET /machines/ response — you don't need to know them in advance. Any field left unspecified falls back to the factory default configuration.

Agents that can only send GET

Some agents' browsing/research mode can only fetch a URL and can't send a POST. recommend and quote therefore also accept GET — just put the parameters in the query string, and it also works pasted straight into a browser:

https://maq.tw/api/v1/recommend?need=我只是 RAG ERP 需求
https://maq.tw/api/v1/recommend?need=本地跑 70B，資料不能外流&budget=800000
https://maq.tw/api/v1/quote?slug=ai-high&graphic=GPU_RTX_PRO_6000_96GB

GET /api/v1 is itself a self-describing endpoint — the get_examples field in its response lists ready-to-use URLs directly. This API has no API key and no User-Agent restriction, and robots.txt is fully open to every AI crawler; if a request fails, check that you're using GET or POST rather than only sending HEAD.

MCP Server

Endpoint https://maq.tw/mcp/　·　Protocol version 2026-07-28　·　Streamable HTTP, POST only

MCP-capable agents (Claude, ChatGPT, Cursor, etc.) can add this URL directly as a server to get six tools:

recommend — pick a machine from a single sentence describing the need, returns a recommended machine and a pre-configured order link
list_machines — machines currently for sale and current pricing
get_machine — specs and configurable options
quote — calculate an exact price for a given configuration
create_quote — generate a short quote URL to hand to the customer (valid for two days)
search_faq — common questions on warranty, after-sales service, loaner units, pre-loaded software, and more

Other discovery channels

llms.txt — a site-wide machine-readable summary, including both endpoints on this page
Every machine's order page carries Product structured data; the FAQ page carries FAQPage structured data

For human readers

If you're a person reading this directly: these endpoints can also be opened straight in a browser — try /api/v1/machines. For custom requirements or partnership inquiries, please contact us via Contact Us, or call (02) 2658-6798.