It looks nice, but after the first use it is not of any help anymore.
So, we have the sizes of the various components, but these remain fixed going forward unless we change Cursor, the extensions, the models or local instructions. Only Conversation will change.
But there is no input/cache/output token explanation, or time breaks for each prompt/reply. Comparing models or prompts brings little information and there is no cost or optimization help here.
Can we expect more from this feature?
Hey, thanks for the specific feedback. Those “what’s missing at first glance” notes are exactly what’s useful.
Basically, the current visualizer is a first iteration. The focus was on showing the categorical breakdown of context system prompt, rules, tools, MCP, conversation, etc. What you’re describing per turn input cache output breakdown, timings, cost optimization hints, model comparisons is the next level, more like an agent usage profiler. Similar requests are already in other threads, for example Context Window Inspector & Agent Usage Profiler and Feature Request: Input Token Estimation. It’s worth adding a vote or comment there so the signal gets aggregated in one place.
If anything moves in that direction, it’ll be in the changelog.
