Zepto Cafe uses MCP to parse text orders, Playwright runs the clicks
Zepto Cafe has taken a modest step toward turning plain‑text requests into a full‑fledged ordering flow, and it does so with two open‑source tools that most developers already know. The project, announced under the banner “Zepto Cafe Now Lets You Order via AI Models,” plugs a lightweight command‑parsing layer into the existing web interface, then hands off the heavy lifting to a browser‑automation framework. By wiring a natural‑language front end to a real browser, the team sidesteps the need for a bespoke checkout API while still letting users say things like “I want a latte delivered to 123 Main.” The approach is deliberately pragmatic: a text‑to‑action mapper reads the instruction, then a separate component actually clicks the buttons, fills in the address and confirms the order.
It’s an example of stitching together off‑the‑shelf pieces rather than building a monolithic system from scratch. The mechanics of that handoff are explained in the following quote.
In Zepto's setup, MCP interprets the user's text instruction and routes it to the appropriate action. Execution of those actions is handled by Playwright, which serves as the browser automation layer. Playwright controls a real web browser to navigate Zepto's website, select the delivery address, add items to the cart, and place the order--replicating the steps a human user would take rather than calling a backend API. In a demo video, Pranav Chandra Prodduturi, a senior category manager at Zepto, showed Claude placing a Zepto Cafe dessert order at a chosen address via the MCP server.
Zepto’s new ordering interface is now publicly available on GitHub. Built by a Zepto engineer, the tool lets users type natural‑language instructions that an LLM—such as Anthropic’s Claude—interprets. MCP, the open‑source Model Context Protocol, acts as the coordination layer, deciding which backend action to invoke.
Playwright then drives a real browser, clicking through the site to fill address fields, add items and confirm payment. The co‑founder and CEO highlighted the demo on LinkedIn, emphasizing that the code runs end‑to‑end without manual integration.
Yet several questions remain. How reliably can the system parse ambiguous requests, and will the browser automation keep pace with peak traffic? The article does not detail error handling or performance metrics, leaving scalability uncertain.
Moreover, the reliance on a third‑party LLM raises concerns about latency and cost that are not addressed. The open‑source nature invites community scrutiny, but practical adoption will depend on how Zepto integrates the prototype into its production stack. At present, the approach demonstrates a functional proof‑of‑concept rather than a finished ordering solution.
Further Reading
- Zepto Cafe MCP: From Natural Language to Automated Orders with Playwright - GitHub
- Playwright MCP: AI-Powered Browser Automation in 2025 - Codemify
- Playwright MCP Reviews and Alternatives 2025 - Skyvern
- Playwright MCP Server 2025: Features & Setup Guide - RanknCompare
- AI Goes Shopping for You with Playwright's MCP Server! - YouTube
Common Questions Answered
How does Zepto Cafe use MCP to interpret user text instructions?
MCP, the open‑source Model Context Protocol, receives the natural‑language request typed by the user and parses it into a structured command. It then decides which backend action—such as selecting an address or adding an item—should be invoked, acting as the coordination layer between the LLM and the automation framework.
What role does Playwright play in Zepto Cafe's AI‑driven ordering flow?
Playwright serves as the browser‑automation layer that controls a real web browser to perform the actions decided by MCP. It navigates Zepto's website, fills in delivery details, adds menu items to the cart, and confirms payment, mimicking the exact steps a human user would take.
Which large language model is used to parse natural‑language orders in the Zepto Cafe demo?
The demo utilizes Anthropic’s Claude as the LLM that interprets the user's plain‑text instructions. Claude processes the request, generates the appropriate command format, and passes it to MCP for routing to the automation stage.
Why does Zepto Cafe choose browser automation over calling a backend API for order placement?
By driving a real browser with Playwright, Zepto Cafe avoids the need to expose or maintain a dedicated ordering API, reducing integration complexity. This approach also ensures the system works with the existing web interface without modifying server‑side code.
Where can developers access the code for Zepto Cafe's AI ordering interface?
The complete implementation is publicly available on GitHub, where the Zepto engineer has published the source code and setup instructions. Developers can clone the repository to experiment with the natural‑language front end, MCP coordination, and Playwright automation.