explaingit

treevu-ai/cli-market-world

2PythonAudience · developerComplexity · 3/5ActiveLicenseSetup · moderate

TLDR

Single CLI, REST API, and MCP server for AI agents to search, compare, and checkout across thousands of VTEX-based retailers, with a parallel price collector.

Mindmap

mindmap
  root((cli-market-world))
    Inputs
      Product queries
      Cart actions
      Retailer logins
    Outputs
      Search and compare results
      Checkout and orders
      Price snapshots in SQLite
    Use Cases
      AI agent comparison shopping
      MCP tool for Claude or Cursor
      Price tracking across VTEX stores
    Tech Stack
      Python
      MCP
      REST API
      SQLite

Things people build with this

USE CASE 1

Let an AI agent comparison-shop across thousands of VTEX-based retailers from one API

USE CASE 2

Run market search, compare, add, and checkout commands from the terminal

USE CASE 3

Wire the 12 MCP tools into Claude, Cursor, or DeepSeek for shopping workflows

USE CASE 4

Collect price snapshots for 40 reference products across many retailers into SQLite

Tech stack

PythonMCPRESTSQLite

Getting it running

Difficulty · moderate Time to first run · 30min

Install with pip and start a local backend; agent and MCP flows need a compatible client like Claude, Cursor, or DeepSeek configured to hit the server.

MIT license covers the code, but the collected pricing data and SKU mappings are kept proprietary by the authors.

In plain English

CLI Market is described as commerce infrastructure for AI agents. The README says that around 3,760 online retailers, including names like Nike, Carrefour, Samsung, and Motorola, all run on a platform called VTEX that exposes the same public API. The problem the project tries to solve is that each retailer still has its own login, its own search, and its own cart, so an AI agent that wants to comparison-shop ends up stuck before its first query. CLI Market puts a single API in front of all of them. You install it with pip, start a local backend, and then use commands like market search, market compare, market add, and market checkout from the terminal. There is a Spanish-language flavor to the examples in the README, with queries for milk, oil, and rice and a payment method called yape, which is common in Peru. The same tool offers an agent mode where you type a natural request such as buy rice and let the system handle the steps, and a JSON output mode for scripts. For AI assistants, the project exposes 12 tools through an MCP server, which is a small protocol used by clients like Claude, Cursor, and DeepSeek. Those tools mirror the CLI commands, covering login, search, compare, cart management, checkout, orders, and reorder. There is also a REST API hosted on a public URL with a Swagger page and an llms.txt file aimed at agents. A second component called the data moat is a price collector that hits 40 reference products across all the retailers every four hours and stores the results in a SQLite database. It runs up to 50 stores in parallel, applies a circuit breaker on repeated failures, and keeps tables for price snapshots, collector runs, and store health. The software is MIT licensed, while the collected pricing data and SKU mappings are kept as proprietary.

Copy-paste prompts

Prompt 1
Walk me through installing cli-market-world with pip and running market search for a product
Prompt 2
Show me how the MCP server exposes the 12 tools that mirror the CLI commands
Prompt 3
Help me wire CLI Market into Claude or Cursor as an MCP server
Prompt 4
Explain how the data moat collector hits 50 stores in parallel and applies a circuit breaker
Prompt 5
Use the JSON output mode of CLI Market to write a script that compares prices across retailers
Open on GitHub → Explain another repo

Generated 2026-05-22 · Model: sonnet-4-6 · Verify against the repo before relying on details.