> ## Documentation Index > Fetch the complete documentation index at: https://docs.kontext.security/llms.txt > Use this file to discover all available pages before exploring further. # Cost > See the tokens and API-equivalent cost of the model requests behind your agents' tool calls. **Cost** shows how many tokens your agents used on tool-related model requests and what those requests would cost at standard API prices. [How cost is calculated](/cost/how-cost-is-calculated) explains which requests count. ## Before you start * You need organization read access in the workspace. Without it, the dashboard shows "You need organization read access to view tool costs." * Endpoints need a CLI release with usage capture. Older releases send tool calls to **Explore** but no token counts. * On the Free plan, Kontext locks **Cost** and other data pages when you use all included monthly transactions. Access resumes when you upgrade or the next cycle starts. ## Read the totals Choose a time range at the top of the page and select **Refresh** to load new activity. The top card shows totals and trend charts. | Control | What it shows | | -------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------ | | **All providers**, **Anthropic**, **OpenAI** | You can view totals and breakdowns for all providers or one provider. Anthropic covers Claude Code and Claude Cowork. OpenAI covers Codex. | | **Total tokens** | Uncached input, cache reads, cache writes, and output of the captured requests. | | **API-equivalent cost** | The same requests priced at standard API prices. | | **Per day** or **Per interval** | Usage within each interval. | | **Cumulative** | Running totals from the start of the range, with a forecast. | | **Model requests** | Captured requests. One request can call several tools. | | **Tool calls** | Distinct tool calls linked to those requests. | | **Show details** | Tokens and API cost for each interval and provider. The dashboard marks intervals that extend beyond the selected range as **Partial**. | Ranges of two days or more use daily intervals in UTC. Ranges from six hours to two days use 24 intervals, and shorter ranges use 48, both in your local time. ## Breakdowns | Section | How it groups requests | | ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Cost by tool activity** | By kind of tool: Shell, Files, Code execution, Computer & browser, Web & network, User interaction, Agent coordination, Timing, Tool discovery. MCP and custom tools appear as `Tool · `. **Mixed tools** holds requests that called several kinds, and **Unknown tool** holds requests with no tool name. | | **Token usage breakdown** | By token type, with the API cost of each. | | **Models** | By model, with requests, tokens, API cost, and the number of unpriced requests. | | **People** | By the identity recorded for the session, or the user assigned to the endpoint. Requests with neither appear as **Unassigned user**. | Each breakdown groups the same requests. Adding two breakdowns counts requests twice. The dashboard warns when requests lack prices or complete token counts. Totals cover only priced requests and reported tokens. If Kontext captured no requests in the range, the dashboard shows **Start with a tool call**. The footer links to each provider's price list and shows when Kontext checked it. ## Forecast Switch to **Cumulative** to see a forecast after the dashed divider. The caption under each chart names the method. | Label | Meaning | | ------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------- | | **Recent average · limited activity** | Extends the average of the latest complete intervals, with no ranges. Kontext uses it when history is short or activity is sparse. | | **Trend forecast · estimated ranges** | Follows the recent trend. Shaded bands show 50%, 80%, and 95% ranges. | Kontext estimates these ranges from limited history and does not guarantee them. The forecast excludes daily and weekly patterns and activity Kontext did not capture. When a forecast is missing, the chart caption gives the reason. Token forecasts need complete token counts. Cost forecasts need a price for every captured request. Both need at least one complete interval after activity starts.