Why We Deleted Our MCP Server and Rebuilt It — Abhi Arya, Reducto

One MCP tool per API endpoint failed once a whole team used it. Fewer tools that carry the structure of real work, plus visible confidence scores, made the agent explainable and trusted.

• Reducto's first MCP server exposed one tool per API endpoint. It demoed well but broke the day the whole team used it: the agent assembled entire workflows confidently and nobody could explain them.
• Abhi Arya deleted that server in one commit and rebuilt it with fewer tools shaped around real work, plus a snapshot tool that gives the agent live system state.
• Arya frames agent-first design around one question, who pays when the agent is wrong, and contrasts three approaches: auto mode, user-first, and agent experience in service of the user.
• The rebuild has the agent ask instead of guess when uncertain and adds instrumentation so usage feeds a self-improving loop. Per the talk, Reducto's sales team then adopted it without being asked.

• MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
• AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.

• videoBrains vs Hands: How to Run AI Agents Safely in Production — Viren Baraiya
• videoWe Mapped 115 Microservices for Our Coding Agents — Kamalakannan Nandagopal, Postman
• videoTurning Agent Memory Into Skills That Work — Will Lyon, Neo4j
• videoLet Anyone at Your Company Ship Internal Apps with AI — Garrett Galow, WorkOS

• videoMCP Doesn't Suck. Your Agent Does. — Jan Čurn, ApifyAI Engineer
• videoRebuilding the web for agents — Liad Yosef, MCP AppsAI Engineer
• videoDashboards Are Dead — Sarah Simionescu, ComposioAI Engineer

Checking sign-in…

Loading comments…