The delivery giant wants iPhone users to order dinner by texting a bot, no app required
DoorDash has opened a waitlist for US iPhone users to test a new AI agent that lives inside Apple’s Messages app. It handles ordering, personalization and checkout without ever opening the DoorDash app.
How the text-ordering agent works #
The feature was unveiled on September 30, 2026, at DoorDash’s Dash Forward event. It is limited to US customers on iOS for now, with Android support planned.
The agent uses your phone number to recognize who you are. From there, it pulls your order history to tailor suggestions.
DoorDash’s own example is a request like “order my usual protein bowl to the office.” The agent is supposed to parse that, figure out which bowl and which office, and build the order.
It can also scan menus, suggest carts complete with images and finish checkout inside the message thread.
Access is tight at the start. Roughly 20,000 consumers already have the preview as part of a pilot, while everyone else can sign up for the broader waitlist at doordash.com/text.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The fine print on AI carts #
DoorDash is being upfront that the agent is not flawless. The company advises users to check AI-generated carts before checking out, since mistakes are possible.
A bigger push into agentic AI #
The Messages agent is not DoorDash’s first AI experiment this year. In June 2026, the company introduced Ask DoorDash, an in-app chatbot built to sharpen ordering recommendations.
Alongside the consumer launch, DoorDash also rolled out an MCP connector aimed at businesses. MCP refers to a standard that lets AI assistants plug into outside services.
The connector is designed for team ordering through corporate systems such as Slack. Early partners include SpaceXAI and Vercel. Companies interested in the connector can join a separate waitlist at developer.doordash.com/mcp.
Co-founder Andy Fang framed the strategy as building AI features that can hold their own against the wave of agentic tools now emerging.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our