Using an MCP to perform product optimizations
Article Summary
Antoine van der Lee stopped guessing what to build next. By connecting his AI coding agent to Amplitude analytics via MCP, he now makes product decisions backed by real user data without leaving his IDE.
This article explores how Model Context Protocol (MCP) lets AI agents access both your codebase and analytics data simultaneously. Antoine shares how he uses Amplitude's MCP with his iOS app RocketSim to optimize features, improve tracking, and focus on high-impact work based on actual user behavior.
Key Takeaways
- MCP is a standardized interface connecting AI assistants to external tools and data
- Agents can query analytics data to validate feature ideas before you build them
- Combining analytics MCPs with Agent Skills creates data-driven optimization plans
- Agents can audit tracking gaps by comparing code against actual analytics events
Connecting your AI coding agent to analytics via MCP transforms product decisions from gut feelings into data-driven choices, all without context switching.
About This Article
Antoine van der Lee was spending too much time manually checking analytics charts in Amplitude. He'd open browsers, create dashboards, and maintain them while working on RocketSim. This slowed him down even though he knew product optimization well from his time at WeTransfer.
He connected Amplitude's MCP server to expose analytics capabilities. This let his AI agents query production tracking data directly and compare it against the codebase. The agents could spot tracking gaps and find logging problems.
His planning prompts got better because they had real data behind them. The agents found that his upsell implementation in BecomeProThumbnailView.swift wasn't working. They gave him direct feedback with clear reasoning on how to increase recording feature conversion among free users.
