An AI agent can now send your mother flowers without you lifting a finger. A new Model Context Protocol server lets language models trigger real-world flower deliveries, part of a growing push to give AI agents the ability to interact with physical services beyond code and text.
From family assistant to flower delivery
Fabian, the developer behind the project, built the MCP while working on hermo.ai, a family AI assistant. The assistant needed a way to trigger tangible actions in the real world, not just answer questions or generate text. Flower delivery became the first concrete use case.
The service, hosted at preznt.net, connects directly to AI agents through the Model Context Protocol, the standard Anthropic introduced for linking language models to external tools and data sources. An agent can place an order without the user creating an account, lowering the barrier for the kind of quick, spontaneous gesture a flower delivery represents.
Currently the service offers a single bouquet type priced at $100 in the United States, 100 pounds in the United Kingdom, and 100 euros in Germany, Switzerland, and Italy. Fabian says the plan is to expand the catalog to include chocolates, gift boxes, and other items that an agent could order directly through the same MCP interface.
MCPs as the bridge to physical goods
The Model Context Protocol has gained traction since Anthropic released it as an open standard earlier this year. Developers have built MCPs for everything from database access to file system operations, but sending flowers represents a different category entirely: a physical good delivered to a real address based on a conversational prompt.
The technical requirement is straightforward. The MCP exposes a tool that an agent can call with recipient details, a delivery address, and payment information. The service handles the rest, routing the order to a local florist network. No account creation means the friction between an agent deciding to send flowers and the order actually going out stays minimal.
That simplicity matters for the use case Fabian describes. A family assistant recognizing a birthday, or a business agent noting a contract renewal, needs to act on that signal without dragging the user through a multi-step checkout flow. The agent makes the decision, the MCP executes it, and the recipient gets flowers.
What the HN community thinks
The project received modest attention on Hacker News, accumulating 8 points and a single comment within three hours of posting. The low engagement reflects both the niche nature of the use case and the early stage of MCP adoption for physical goods. Most MCP discussions on the platform center on developer tooling, not consumer-facing delivery services.
That said, the underlying idea touches on something the AI agent community has been circling: the gap between what agents can reason about and what they can actually do in the world. An agent that can draft a thoughtful message about a customer renewal is useful. An agent that can also send flowers to mark the occasion closes a loop that most tools leave open.
Expanding the catalog and the model
Fabian's roadmap includes adding more product categories, chocolates and gift boxes being the most natural next steps. Each additional category requires connecting to a different supplier network, but the MCP interface stays the same: the agent calls a tool, the service handles fulfillment.
The question for adoption is whether developers building AI assistants find enough value in the flower delivery use case to integrate it. For a family assistant like hermo.ai, the connection is obvious. For general-purpose coding agents or enterprise tools, the relevance depends on whether the agent operates in a context where sending physical goods makes sense.
For now, the project sits at the intersection of two trends: AI agents gaining the ability to act in the real world, and MCP servers becoming the glue that connects those agents to services beyond the screen. Flowers are a small, specific example, but the pattern they establish could extend to any service that an agent might need to trigger on a user's behalf.