DoorDash launches a limited beta of DoorDash CLI, which lets users place orders via an AI agent, available by waitlist to macOS developers in the US and Canada
Sudo make me a sandwich. The future has arrived! DoorDash just introduced a limited beta of DoorDash CLI …
TechCrunchSarah Perez
Context & Ripple Effects
DoorDash has been extending AI across several ordering surfaces: restaurant voice ordering in 2023 and, more recently, Ask DoorDash for food, grocery and reservation requests made with prompts or photos. The CLI beta moves that experimentation from an in-app assistant to a developer-facing agent interface.
The rollout is narrowly scoped—a waitlist for macOS developers in the US and Canada—but it tests whether ordering can become an action delegated to software agents rather than a task completed inside DoorDash’s consumer app.
First-order effects
Developers admitted to the beta can have an AI agent place DoorDash orders through the CLI, creating a new ordering entry point alongside DoorDash’s app-based conversational tools.
DoorDash gains a controlled environment to evaluate agent-driven ordering workflows before making the capability broadly available to consumers or other platforms.
Second-order effects
Restaurants and merchants connected to DoorDash may receive orders originating from agent workflows, increasing the importance of accurate menus, availability and order-handling rules that an agent can interpret reliably.
Other delivery platforms and ordering providers face pressure to offer agent-compatible interfaces if developers begin treating food and grocery delivery as a callable service rather than a destination app.
Third-order effects
If agent-mediated ordering expands beyond a developer beta, delivery marketplaces could compete increasingly on the quality, reliability and governance of their machine-action interfaces—not only consumer app discovery.
The pattern points toward AI assistants consolidating routine purchasing decisions; that could shift marketplace leverage toward the platforms that control both merchant fulfillment and the agent integration layer, though adoption will depend on trust and execution reliability.
The trend: DoorDash CLI is part of the broader shift from AI that helps users search and compose requests to AI agents that can carry out transactions across commerce services.
Today we're opening up the DoorDash CLI in limited beta. ‘dd-cli’ lets you order DoorDash directly from your agent: search stores, find the best deals, check out, and more. Early access for US/Canadian macOS developers by waitlist. Excited to see what folks build! [video]
got excited at first, but then realized it kind of sucks for our use case: - darwin-arm64 binary only, no linux build - login needs a local browser + localhost redirect, no headless/device-code auth - access is per-account waitlist, so my users can't bring their own
so this is a big milestone right now these companies are in a delicate balance - they own the end relationship with the customer and being disintermediated would be a bad thing but if there is a real paradigm shift happening then they're dead if they don't offer this
i hope this is what the rest of 2026 looks like; companies opening up cli access and in 2027 i want my robot to stop folding laundry and go get the food at the door and set the table
A CLI means you can drop DoorDash into whatever you already run. Wire it into your internal tools and office catering orders itself. Point an expense agent at it and DoorDash receipts pull themselves. Give an assistant your calendar and lunch shows up in the gap between
I'm v surprised to see DoorDash offer this b/c low-margin businesses (food delivery is low margin!) NEED to own the end-to-end customer relationship. That's how they eg sell ads (that makes them profitable) It's why these services eg never offer APIs for devs to build own apps!
doordash having a CLI feels stupid for about 5 seconds and then you realize every consumer app is probably going to need some version of this for agents kind of funny that browser use was supposed to be the end game and we're now back to building a separate interface for every
Great work @andyfang. This marks a profound shift in inference—moving from humans querying AI, to AI orchestrating human action in the physical world. Starting with delivery, expanding to an entirely new agentic economy.