DoorDash's New AI Tool Lets You Order Food from the Command Line
DoorDash is testing a new way to order food that might seem like a programmer's inside joke—but it's real. The company has launched a limited beta of DoorDash CLI (dd-cli), an AI-powered command-line tool that lets developers place food delivery orders directly through an AI assistant. Yes, you read that right: you can now order a burrito using the same interface you use to run code.
What is dd-cli?
DoorDash CLI is exactly what it sounds like: a command-line interface for DoorDash. But instead of typing git commit or npm install, you can ask an AI assistant to search for nearby restaurants, find promotions, and complete checkout—all without ever opening the DoorDash app. The tool is currently available to macOS developers in the United States and Canada, but only through a waitlist.
Andy Fang, DoorDash's co-founder and CTO, announced the new feature on X (formerly Twitter). While command-line tools are typically reserved for software development, using one to order lunch might seem like a joke. But Fang and his team are serious. They see dd-cli as an exploration of what they call "agentic commerce"—where AI agents act on behalf of users to complete commercial tasks.
Why would anyone order food from a terminal?
It's a fair question. For most people, opening the DoorDash app is faster and easier than typing commands. But the point isn't to replace the app for everyday users. Instead, DoorDash is opening up its ordering capabilities so that AI agents—like ChatGPT, Claude, or custom-built assistants—can call DoorDash's service APIs. This means developers can integrate food ordering, grocery shopping, and local promotion searches into their own software, smart assistants, or other services.
In other words, DoorDash wants to be part of the AI ecosystem. Instead of forcing users to open an app, they're betting that AI agents will become a new entry point for connecting consumers with commercial services. Imagine telling your AI assistant, "Order me a salad from the place I like," and it handles everything—from finding the restaurant to applying coupons to placing the order.
How it works (and why it's more complex than you'd think)
In a demonstration video, the AI assistant reads Slack messages, parses JSON data, checks menu structures, runs Python scripts, and adjusts operations based on error feedback—all to complete multiple salad orders. It's much more complex than regular ordering, but that's the point. The process shows how AI agents can autonomously call tools and execute tasks, even when things go wrong.
This isn't DoorDash's first foray into AI-powered ordering. The company previously offered meal ordering via iMessage and launched an AI chatbot called "Ask DoorDash." It also opened up its capabilities to third-party AI assistants like OpenAI's ChatGPT and Anthropic's Claude. The dd-cli beta is the latest step in building an AI-driven consumer service ecosystem.
What this means for the future of commerce
As AI assistants move from simple information retrieval to task execution, companies like DoorDash are trying to integrate commercial services directly into the AI ecosystem. The vision is a shift from "users actively operating apps" to "AI agents handling services on behalf of users." If that sounds like a sci-fi future, it's closer than you think.
For now, dd-cli is just a limited beta. But it's a glimpse of a world where ordering food, booking a ride, or buying groceries might not require a single tap on a screen. Instead, you'll just ask your AI assistant—and it will take care of the rest.

Key Points
- DoorDash launches limited beta of dd-cli, an AI command-line tool for ordering food via AI assistant.
- Available to macOS developers in US and Canada through a waitlist.
- Tool allows AI agents to search merchants, find promotions, and complete checkout.
- Represents DoorDash's exploration of "agentic commerce"—AI agents handling commercial tasks.
- Previous efforts include iMessage ordering, "Ask DoorDash" chatbot, and integrations with ChatGPT and Claude.