DoorDash Launches CLI Tool That Lets AI Agents Order Food From the Command Line

DoorDash has introduced a limited beta of DoorDash CLI, a new command-line tool that allows developers to order food directly through AI agents. Known as dd-cli, the tool can search stores, find deals and complete purchases, giving developers another way to interact with DoorDash’s delivery platform.

The beta is available through a waitlist for macOS developers in the United States and Canada, DoorDash co-founder and CTO Andy Fang announced on X. While the idea of ordering lunch from a command-line interface has generated amusement online, the company positions the tool as a practical example of how AI agents can carry out real-world commerce tasks.

DoorDash CLI gives developers access to the company’s ordering platform, allowing them to build applications that can order food or groceries or discover local deals without relying on the standard DoorDash app. The functionality can also be integrated with other software and AI-powered services to create more advanced automated workflows.

The launch builds on DoorDash’s broader push into agentic commerce. The company has previously experimented with food ordering via iMessage, introduced its own AI chatbot, Ask DoorDash, and already makes its platform accessible to AI assistants, including OpenAI’s ChatGPT and Anthropic’s Claude.

Developers signing up for the beta are asked what they would build using the CLI, suggesting DoorDash hopes the community will create new AI-driven applications beyond traditional food delivery. The announcement also references the classic XKCD comic featuring the phrase “sudo make me a sandwich,” a well-known joke among programmers about using administrator privileges to accomplish a simple task.

DoorDash embraced that humour in its demonstration video, which shows an AI agent reading Slack messages, recalling memories, parsing JSON, analysing menus, running Python scripts, recovering from errors and calculating totals before finally ordering three salads. The intentionally over-engineered process playfully illustrates how AI agents could eventually automate everyday purchasing tasks.