> ## Documentation Index > Fetch the complete documentation index at: https://docs.barndoor.ai/llms.txt > Use this file to discover all available pages before exploring further. # Reading the Dashboard > What the counts, policy decision totals and top-activity tables on the Barndoor home Dashboard measure, and where they differ from the usage dashboards. The **Dashboard** (first item in the sidebar) is the admin home page: headline counts for your organization, policy decisions over time, and the most active users, AI agents, MCP servers and tools. This page defines what each figure counts. The Dashboard reports MCP activity and policy decisions. It has no token or cost figures — for LLM Gateway traffic, use the [LLM Usage Dashboard](/how-tos/read-llm-usage-dashboard), and for MCP call volume, duration and errors, the [MCP Usage Dashboard](/how-tos/read-mcp-usage-dashboard). ## The time period The selector in the top right — **Last 24 Hours**, **Last 7 Days**, **Last 30 Days** (the default), **Last 90 Days**, **Last 6 Months** or **Last 12 Months** — sets the window for everything below the three count cards. Each window is rolling and ends at the current time. Unlike the usage dashboards, the Dashboard groups activity into hours, days and months in **your browser's timezone**, so two admins in different timezones can see different per-day figures for the same window. ## The count cards | Card | What it counts | | - | - | | **AI Agents** | AI agents registered in your organization | | **MCP Servers** | Active MCP servers that you can manage policies for | | **Users** | Users in your organization | These are current totals, not activity. They don't change with the time period, and a server or agent with no traffic still counts. Each card links to its list page. ## Monitored Actions Policy decisions over the window, with three cards above a line chart: | Card | What it counts | | - | - | | **Total Policy Decisions** | Every policy decision recorded in the window | | **Allowed Requests** | Decisions that allowed the call, including calls that needed approval | | **Denied Requests** | Decisions that denied the call | These count **policy decisions, not completed tool calls**. A call that was allowed and then failed on the MCP server still counts as allowed, and a call that needed approval counts as allowed whether or not anyone approved it. Beneath each count is a rate and a comparison, for example `2 / day • 40 in previous 30 days (↑ 50%)`: * **The rate divides by the number of periods that had activity**, not by the length of the window. A 30-day window with activity on only 10 days divides by 10, so the rate reads as "per active day". The unit is hours for **Last 24 Hours**, months for **Last 6 Months** and **Last 12 Months**, and days otherwise. * **The comparison** is against the window of the same length immediately before this one. Selecting a card opens **Audit History** filtered to those decisions over the same window. ## Top Users The ten most active users in the window, with two columns: | Column | What it counts | | - | - | | **Actions** | Audit events recorded for the user on currently active MCP servers | | **MCP Servers Used** | How many of those servers the user was active on, shown as "*n* of *total*" against the **MCP Servers** card | **Actions counts audit events, not tool calls.** A single tool call records more than one event — the policy decision and the call itself, for example — so a user's Actions is typically higher than their number of tool calls. Only servers that are currently active count. Activity on a server that has since been deleted or deactivated is left out of both columns, which keeps **MCP Servers Used** from exceeding the total. Selecting a row opens **Audit History** for that user over the same window. ## Top Usage The five most active **AI Agents** and the five most active **MCP Servers** in the window, each ranked by **Actions** — audit events, as in Top Users. Activity with no recorded agent or server name is left out. Unlike Top Users, this card isn't limited to currently active servers, so a server removed during the window can still appear. ## Tool Calls The ten busiest tools in the window, one row per MCP server and tool: | Column | What it shows | | - | - | | **Type** | **Read**, **Write** or **Unknown** | | **Total Calls** | Policy decisions for the tool | | **Allowed** | Decisions that allowed the call, including calls that needed approval | | **Blocked** | Decisions that denied the call | Like Monitored Actions, the counts are policy decisions. Decisions made by Data Control Center runtime policies are left out. **Type** comes from the tool's own metadata when Barndoor has it. Otherwise it's inferred from the tool's name — names starting with verbs like *create*, *update*, *delete* or *send* read as **Write**, and *get*, *list*, *search* or *fetch* as **Read**. Anything else shows **Unknown** rather than guessing. ## Frequently asked Actions counts audit events, and a tool call usually records more than one. Use **Top Users** to compare users against each other, and the [MCP Usage Dashboard](/how-tos/read-mcp-usage-dashboard) — grouped by **User** — for call counts. It divides by days that had activity, not by every day in the window. A burst of activity on a few days produces a high per-day rate even over a long window. The two count different things. The MCP Usage Dashboard counts MCP messages proxied through the gateway; the Dashboard counts policy decisions (Monitored Actions, Tool Calls) and audit events (Top Users, Top Usage). The Dashboard also groups by your browser's timezone, while the MCP Usage Dashboard defaults to UTC.