Making Magic with MCP: From Data Retrieval to Real Analysis and Insights

Summary

Actionable Intelligence to Improve How the Ably Team Works The real power emerged when I moved beyond simple data retrieval to analysis and insights. When I asked, “How is developer productivity in the xyz team this month?” Something awesome happened. The AI didn’t just return raw metrics. Instead, it: Intelligently identified all key metrics associated with developer productivity Analysed the data to understand potential reasons for performance patterns Connected insights across different Jellyfish data points, linking productivity metrics to investment levels and deliverable timelines Most impressively, it concluded that productivity appeared lower because the team was split across multiple investment categories and several outcomes had just kicked off. This matched our internal hypothesis perfectly, but the AI had arrived at this conclusion independently by analysing the data relationships in seconds. I then tested more specific queries, asking “How does PR cycle time compare within the xyz team for April?” The results went far beyond a simple data dump: Individual PR cycle times were listed for each team member The AI provided contextual explanations for why each person’s metrics might look the way they did It identified patterns, noting that certain team members had higher PR review counts Most remarkably, it correctly inferred these were likely tech leads carrying a heavier review burden This wasn’t just showing data, it was providing actionable intelligence that could be used to improve processes and ways of working. The true potential became apparent when I connected additional MCP servers alongside Jellyfish. Within minutes, we could: Ask Claude for insights from Jellyfish data about team performance Have it analyse our documented processes in Confluence to identify potential inefficiencies Receive specific suggestions for process improvements based on the gap between documented intent and measured reality This connected workflow happened seamlessly in a si...