How technical support at Cursor uses Cursor Cursor

"By collapsing code, logs, team knowledge, and past conversations into a single Cursor session, we've removed that bottleneck for most of our work. Today, over 75% of Cursor's support interactions run through Cursor itself, increasing support engineer throughput 5-10x."

"Because our full codebase is available locally, Cursor can index and use semantic search across product code, docs, and internal tooling in the same session. Product context almost always spans multiple repositories, and we keep related repositories together in one workspace so that kind of question is answerable in a single thread."

"MCP servers allow us to integrate databases with customer information, streamed event logs containing details on services used and telemetry errors, communication platforms like Slack filled with threads and conversations, and engineering ticket platforms containing potentially dozens of unique teams that each operate differently."

Support investigations are fundamentally research problems requiring extensive context gathering. Cursor addresses this bottleneck by integrating code, logs, team knowledge, and past conversations into unified sessions. Over 75% of Cursor's support interactions now run through Cursor itself, dramatically increasing support engineer throughput by 5-10x. Investigations typically begin in Ask Mode, tracing symptoms backward through product behavior using semantic search across codebases, documentation, and internal tools. Multi-root workspaces enable engineers to access related repositories simultaneously, allowing complex questions spanning frontend logic, backend policies, and documentation to be answered in single threads. MCP servers further enhance investigations by integrating customer databases, event logs, communication platforms, and engineering ticket systems, eliminating the need to search across multiple tools.

Read at Cursor