> ## Documentation Index > Fetch the complete documentation index at: https://docs.corepillar.ai/llms.txt > Use this file to discover all available pages before exploring further. # AI Usage > The AI Usage page in ModelHub shows your own spend, requests, tokens, model mix, privacy events and routing savings over the last 7, 30 or 90 days. **AI Usage** in the sidebar shows what your AI use costs and what was protected. For employees it is titled **My AI usage**, with the subtitle "Your activity only". It is the same screen administrators use in AgentHub, run on the same numbers, narrowed to you. You and your administrator always see identical arithmetic for your account. ## Scope * Your view covers **your own activity across every project in the organization**, not only the project you are in. * It includes chat and any other CorePillar feature that called a model on your behalf. * Nobody else's activity is included, and no URL change can widen it. Organization owners, admins and super admins get an extra **Organization | Just me** toggle at the top. **Just me** is the personal view described here; **Organization** is the org-wide dashboard documented in [AI Usage (AgentHub)](/agenthub/ai-usage). Choosing **Just me** adds `?scope=personal` to the URL. ## Time range The **7d / 30d / 90d** control at the top right sets the window. The default is **30d**. The window is a rolling period ending now, and days are grouped by UTC date. Changes from the previous period compare against the same number of days immediately before the window. ## Summary tiles | Tile | What it counts | Change indicator | | - | - | - | | **Total spend** | Dollar cost of every model call in the window | Percentage change against the previous window. A fall is shown in green, a rise in red. | | **Requests** | Model calls across all sources (each chat answer is one) | Percentage change. A rise is shown in green. | | **Tokens** | Input plus output tokens | None | | **PII stripped** | Total personal-data items removed from your messages | None | A dash (—) means there is nothing to compare against. "flat" means the change rounds to 0%. ## Spend over time An area chart of daily spend, with a dashed line for **Requests (relative)**. The request line shares the chart and is scaled to it; it has no axis of its own. Hover a day to see its date, cost and request count. With no activity in the window it shows "No data yet". ## Spend by source A table and a stacked daily bar chart, split by the feature that made the call: | Source | What it covers | | - | - | | **Chat** | ModelHub conversations | | **Cost Calculator** | Cost Calculator runs in AgentHub | | **Playground** | Playground runs | | **Evaluations** | LLM-judge evaluator calls. These are counted as requests; their tokens and cost are not stored, so the row reads "tokens & cost not recorded". | Columns: **Requests**, **Tokens**, **Cost**, **Share** (of total cost). Most employees only have a **Chat** row. ## Tokens in / out A stacked daily bar chart of **Input tokens** and **Output tokens**. Hover a bar for that day's numbers. ## Activity A calendar heatmap of the last 13 weeks, one square per day, shaded from **Less** to **More** by the number of turns. Hover a square for its date, turn count and cost. Days outside the selected range show as empty. ## Model mix A donut of spend by model with **Total spend** in the centre, and a table: | Column | Meaning | | - | - | | **Model** | Display name of the model | | **Requests** | Calls to that model | | **Cost** | Spend on that model | | **Avg latency** | Average total time per chat answer, in milliseconds | | **Right-sized** | Share of chat answers on that model whose verdict was right-sized | ## PII leakage | Panel | Shows | | - | - | | **Events over time** | Daily count of items removed from your messages. A red **N turns blocked** badge counts messages that were held for confirmation because something critical was found. | | **By severity** | A bar split into critical, high, medium and low, with counts | | **By type** | The eight most frequent types (for example `EMAIL`, `PERSON_NAME`, `OPENAI_API_KEY`), with counts | Your view never shows the removed values themselves, only types and counts. With nothing detected, each panel says "No PII detected". ## Routing & savings The header carries three pills with the share of your chat answers in each verdict: **Right sized**, **Overpowered** (shown as **Overkill** on answer cards) and **Underpowered**. | Figure | Meaning | | - | - | | **Saved by auto-routing** | What your Auto answers would have cost on the most expensive model connected to the project, minus what they did cost. Underneath: "N of M turns on Auto · K escalated", where escalated means Auto decided a message needed more than the Small tier. | | **Prompt optimisation** | Estimated tokens removed by [Optimise](/modelhub/chat/optimise), and how many turns it changed | | **Right-sizing headroom** | What you would have saved if every overpowered answer had used the cheaper model suggested for it | | **Tier split** | How many of your messages were scored as needing Small, Mid and Large | ## How the numbers are calculated * **Cost** comes from the provider's reported token counts priced against CorePillar's model price list. * **Chat turns** are recorded when the answer finishes. A stopped or failed answer may not appear. * Older chat turns recorded before savings were re-priced on actual tokens are re-derived from their tokens, so savings are not overstated. * The page loads both your view and, for administrators, the organization view at once, so switching the toggle is instant once both have loaded. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.