Restock
A sample AI procurement agent project built on the Managed Deep Agents framework, integrating Slack, Stripe, and Zinc. It supports natural-language purchase requests, product search, and payment following human approval, serving as a reference implementation demonstrating AI agents executing real-world transactions.
Timeline (last 90 days)
Restock is a sample project built on Managed Deep Agents (MDA); it is an AI procurement agent capable of receiving purchasing requests in Slack, searching for real products using Zinc, making payments via Stripe Link, and completing orders.
Restock adopts a human-in-the-loop design in which the agent does not place orders on its own but instead presents the procurement details to the user for review in Slack before payment is made.
In the Restock workflow, payment is completed via Stripe's Link, and Zinc handles the checkout action on the retailer side.
The developer has open-sourced the Restock sample project, accompanied by a blog post explaining how it works, along with the code and configuration steps.
Restock provides a rehearsal mode that allows users to experience the complete workflow from receiving a request to "ready to pay" without actually placing an order or making a payment.
The design concept of rehearsal mode is similar to a "sandbox environment" or "dry run" in software development, executing all business logic but stopping—or substituting simulated responses—before triggering real charges or order placement.
An "AI prepares, human approves" workflow that retains explicit human approval at high-risk payment nodes is easier to deploy in enterprise finance and compliance scenarios than a fully autonomous spending bot.
Restock is currently positioned as a sample agent rather than a mature commercial product; real deployment still requires consideration of permission management, budget controls, error rollback, and handling mechanisms for when the agent selects the wrong item.
All Facts (8)
Restock is a sample project built on Managed Deep Agents (MDA); it is an AI procurement agent capable of receiving purchasing requests in Slack, searching for real products using Zinc, making payments via Stripe Link, and completing orders.
50%UnverifiedRestock adopts a human-in-the-loop design in which the agent does not place orders on its own but instead presents the procurement details to the user for review in Slack before payment is made.
50%UnverifiedIn the Restock workflow, payment is completed via Stripe's Link, and Zinc handles the checkout action on the retailer side.
50%UnverifiedThe developer has open-sourced the Restock sample project, accompanied by a blog post explaining how it works, along with the code and configuration steps.
50%UnverifiedRestock provides a rehearsal mode that allows users to experience the complete workflow from receiving a request to "ready to pay" without actually placing an order or making a payment.
50%UnverifiedThe design concept of rehearsal mode is similar to a "sandbox environment" or "dry run" in software development, executing all business logic but stopping—or substituting simulated responses—before triggering real charges or order placement.
50%UnverifiedAn "AI prepares, human approves" workflow that retains explicit human approval at high-risk payment nodes is easier to deploy in enterprise finance and compliance scenarios than a fully autonomous spending bot.
50%UnverifiedRestock is currently positioned as a sample agent rather than a mature commercial product; real deployment still requires consideration of permission management, budget controls, error rollback, and handling mechanisms for when the agent selects the wrong item.
50%