This is my own AI coding spend — measured, not marketed.
Every number below is pulled from my real Claude Code and Cursor usage, the same dashboard I run on my own work. No sample data, no rounded-up demo figures — the product, running on itself.
Trailing week across both tools. Hover any tile for the all-time total.
Real cost, by day and by model
Daily spend stacked by model over the last 30days, and the all-time split across Fable, Opus, Sonnet, Haiku, and Cursor’s models.
Everything the work leaves behind
A living vault of notes, decisions, and entities — folder-level counts and growth only, never titles or contents.
Notes by folder
Top-level folders in the vault
Vault growth
Total notes tracked over time (history starts when tracking began)
Second brain
Every dot is a real, captured decision or piece of project context the agent can pull from instead of re-deriving it next session. It's tracked separately from the cost multiple above — not folded into that number — but a knowledge base this connected is part of why that multiple holds up as work gets more complex, not just cheaper on day one. Folder categories and link structure only; hover a node for its category, never a title or contents.
Want this rolled up across your whole team?
This page is one developer’s usage. Costero brings the same cost, efficiency, and adoption visibility to entire teams and orgs — per person, per repo, org-wide.