Agentic commerce: when AI agents stop chatting and start transacting
Agentic commerce is the shift from AI as a conversational helper to AI agents that directly use APIs and CLI tools to execute real-world transactions—such as autonomous food ordering or domain management—on a user’s behalf without needing manual clicks in a traditional app. In other words, software stops being a front-end for humans and becomes a contracting layer where agents search, compare, and purchase services as programmable steps in a workflow. That transition changes who owns the customer relationship, how safeguards must be designed, and which companies stay visible in autonomous workflows that may never show a branded interface again.
DoorDash and GoDaddy have now put a stake in the ground: if agents are going to run people’s day-to-day tasks, those agents need direct access to commercial services. DoorDash’s new dd-cli beta lets AI agents order food and groceries from the command line, while GoDaddy’s developer platform exposes registrar APIs and a CLI so agents can handle search, purchase, and DNS configuration inside code-driven workflows. The message is clear—if your service cannot be called by an agent, it might as well not exist in an automated world.
DoorDash’s dd-cli: autonomous food ordering as a terminal command
DoorDash has launched the dd-cli beta, a command-line tool that AI agents can use to search stores, compare prices, and order groceries. Andy Fang, the company’s co-founder and CTO, announced that the CLI is available via waitlist to macOS developers in the US and Canada, opening the door for early adopters to wire DoorDash directly into their agents. Practically, that means a developer can create a personal assistant that searches nearby restaurants, compares prices, builds a cart, and pays with a single command, no app screen or human review in the loop.
This is not DoorDash’s first AI integration—they already have a Claude connector, a ChatGPT integration, and an in-app assistant where humans still approve the final order. The dd-cli is different because it assumes agents can be trusted to act autonomously. The waitlist itself requires a social link and a short explanation of why you want this power, which is a soft guardrail but also a signal: DoorDash knows it is giving up control over its carefully tuned app flows in exchange for participating in autonomous food ordering ecosystems.
GoDaddy’s domain APIs: domains as programmable infrastructure for agents
On the other side of the stack, GoDaddy has launched a new developer platform that brings domain management into the development environment instead of a web dashboard. The platform exposes domain APIs that allow agents and applications to search for available domains, retrieve pricing, purchase names, configure DNS records, and manage registrations without ever leaving the workflow. There is also a CLI for those who prefer command-line control, mirroring the pattern emerging at DoorDash—CLI tools for AI agents that can be scripted straight into CI/CD or Infrastructure as Code pipelines.
GoDaddy’s bet is that domains are becoming another programmable component of application infrastructure, and AI is reshaping how software is created. By making registration and purchasing programmable, they go beyond cloud DNS providers that mainly handle records for domains registered elsewhere. The practical impact for ordinary users is subtle but real: a domain management agent can now spin up a domain, wire DNS, and ship a site in minutes as part of a single automated workflow. That turns what used to be a separate registrar chore into a background operation that a domain management agent quietly handles while you focus on building the product.
Guardrails and the uncomfortable trade-off: control versus relevance
Once agents can move money and change infrastructure through AI agents APIs, guardrails stop being optional. GoDaddy built its v3 Domains API around a quote-then-execute model, using a short-lived token to lock in price, an idempotency key to prevent duplicate charges, and a consent object that ties each registration to human approval. The platform uses OAuth-based authentication with scoped permissions so AI agents only get access to specific resources, and irreversible actions are wrapped in approval flows to keep user control.
DoorDash’s guardrails are earlier-stage and more social than technical—limited beta access, macOS-only, and a waitlist that asks why you need an agent ordering burritos at 3am. But the deeper tension is about customer ownership. Critics point out that delivery platforms make thin margins and rely on owning the full end-to-end relationship, including advertising and sticky app flows. When an agent fires off a dd-cli command, it skips all that. Yet as one observer bluntly put it, “If there is a real paradigm shift happening, then they’re dead if they don’t offer this”. Control is being traded for relevance in agent-led ecosystems.
Competitive necessity and the future of agentic commerce
As AI agents evolve beyond writing code toward handling everyday tasks, infrastructure has to keep pace. DoorDash’s dd-cli and GoDaddy’s domain management agents platform are not isolated experiments; they are early signs of how agentic commerce will work. Agents will expect programmable access to food delivery, domains, logistics, payments, and everything in between. If they cannot call you, they will route around you—to a competitor or to a hacked-together workaround—and your carefully designed app UX will never even be rendered.
For engineering teams, the takeaway is that domains and delivery services are becoming programmable building blocks, not separate destinations. For business leaders, the choice is starker: offer sanctioned agent access with strong guardrails, or risk obsolescence as autonomous workflows consolidate around the platforms that do. GoDaddy is already co-authoring an IETF draft for an Agent Name Service standard, aiming to make agent identity and verification part of the internet’s fabric. That is the trajectory—AI agents will not only talk to users, they will transact on their behalf. Companies that accept that and design for it will shape the next phase of online commerce; those that stall will find their brands quietly fading from the agent’s field of view.
