Dashboard

    The ManageEngine SecureAI dashboard brings the AI agents and Model Context Protocol (MCP) servers in your organization together into one view. It offers:

    • An inventory of your AI agents: Displays agent counts by platform, guardrail coverage, the Enterprise AI Risk Score, and the MCP inventory.
    • A record of tool call activity: Displays allowed, blocked, and redacted tool calls, their latency, and their origin.
    • Quick access to pending work: Displays the workflow requests awaiting your action, and shortcuts to the records behind each metric.

    Use the dashboard to identify items that need attention, such as an agent with no owner, a platform with low guardrail coverage, or an increase in blocked tool calls. Select a metric to view the records behind it. You can then assign owners, apply guardrails, adjust control policies, or correct gateway connections.

    How the dashboard works

    The dashboard stores no data of its own. Each time you access it, every widget counts records that SecureAI has already gathered from two sources.

    The agent counts, guardrail coverage, and AI inventory figures come from the scheduled sync with your platform accounts. SecureAI connects to each onboarded account and reads what it exposes through its APIs: agents, models, knowledge bases, guardrails, and templates, each stored with its platform, model, owner, and status. These figures therefore show the position as of the last sync, and an agent created after that sync does not appear until the next one completes.

    The tool call, latency, and location figures come from the gateway. Agents reach MCP servers only through an MCP gateway, so every call passes through a point where SecureAI can observe it. As each call completes, the gateway logs an entry with the caller's identity, agent, MCP server, tool, timestamp, latency, matched policies, outcome, source IP address, and the request and response bodies. SecureAI resolves the location figures from that IP address.

    You see only the widgets that your delegated role permits. The Enterprise AI Risk Score is the one figure the dashboard does not count. SecureAI takes it from the most recent risk assessment, and it changes only when a new assessment finishes.

    Prerequisites

    • Log in with the necessary role delegated to your technician account. You see only the figures your role covers, so if you hold only one of these help desk roles you will find parts of the dashboard empty.
    • You must have a workflow role for the stage whose pending requests you want to act on: Reviewer for Review, Approver for Approval, or Executor for Execution.
    • Ensure there is at least one onboarded agent platform instance, with its first sync complete.
    • Ensure there is at least one MCP gateway, with at least one registered MCP server showing as connected. Tool calls are logged only when they pass through a gateway.
    • Have the control policies, data loss prevention (DLP) policies, and guardrails you want reported on.
    Notes:
    • The dashboard is for review only. You cannot add, remove, resize, or rearrange widgets, and you cannot save a personalized layout.
    • You cannot change the period that any widget covers. The charts show a fixed rolling window of the last seven days, and the remaining widgets show current totals.
    • You cannot export, print, or schedule a widget from the dashboard. Use the corresponding report instead.
    • A newly created agent does not appear on the dashboard until the next sync completes.
    • Most Used MCP Servers, Agents by Platform, and Guardrail Coverage show only the top entries or the platforms that returned data. Use the corresponding report for the complete list.
    • A widget with no data shows an empty state instead of a zero.

    Configuration instructions

    1. Open the Dashboard tab. The General tab opens by default.
    2. Review the summary counts:
      Summary What it shows
      Agents Total, privileged, and unmanaged agent counts. Click Unmanaged to open the Unmanaged Agents tab, or Privileged to open the Privileged Agents tab.
      MCP Servers Total, protected, and inactive server counts.
      Agents Without Guardrails The number of agents with no guardrail applied. Select the count to list them.
      Enterprise AI Risk Score The score and band from the last risk assessment. Click Analyze Now to run a new assessment.

      It also provides the following widgets, which break those counts down: Agents by Platform, Guardrail Coverage, MCP Gateway Activity, Most Used MCP Servers, and AI Inventory.

      Widget What it shows
      Agents by Platform The total agent count, split by platform: Copilot Studio, Zoho Zia Agents, and Gemini Enterprise Agent Platform. Select a legend entry to view the agent report for that platform.
      Guardrail Coverage The percentage of agents that have a guardrail applied, with the counts behind it.
      MCP Gateway Activity Recent tool calls, updated as new calls are logged. Each entry shows the tool name, MCP server, the outcome (Allowed, Blocked, or Redacted), the timestamp, the matched policy, the source IP address, and the MCP gateway. Click an entry to view the Gateway Activity details, or click the refresh icon to reload the list.
      Most Used MCP Servers MCP servers ranked by the number of tool calls they received.
      AI Inventory Counts of MCP Servers, Registered MCP Tools, Knowledge Bases, Guardrails, Model Armor Templates, Action Group Lambda Functions, and Identity Blueprints.
    3. Open the Access Protection tab, then review its summary counts:
      Summary What it shows
      MCP Configurations The number of MCP servers and gateways in your setup.
      Protection Overview Tool call and DLP redaction counts.
      Control Policies Access and DLP policy counts.

      Review its widgets: Tool Call Outcomes, Tool Calls Blocked by Control Policy, Tool Call Latency Trend, and Tool Calls by Location.

      Widget What it shows
      Tool Call Outcomes The total number of tool calls, split into allowed and blocked. Redacted calls are counted as allowed. Select a legend entry to list the calls behind it.
      Tool Calls Blocked by Control Policy The number of tool calls blocked by a control policy on each day of the rolling window.
      Tool Call Latency Trend Average tool call latency, in milliseconds, across the rolling window.
      Tool Calls by Location The number of tool calls originating from each country.
    4. Use the side panel, which stays available in either view:
      • Pending Requests shows the number of workflow requests awaiting your action, split across Review, Approval, and Execution. Select a count to view that queue.
      • The Governance, Access Protection, and Reports sections link to frequently used pages.
    5. To assign owners to agents that have none, click the Unmanaged count under Agents, select the agents, then click Assign Owner.
    6. To investigate an individual tool call, click any entry in the MCP Gateway Activity widget. Expand Arguments to see the parameters the agent sent, and Response to see the result returned. If the response was redacted, click View Redaction Details to see what was masked and by which policy.

    Tips

    • Assign owners to unmanaged agents. An agent with no owner has nobody to contact when something goes wrong, and it is skipped by access certification.
    • Protect every MCP server—an unprotected server can be reached without a policy check.
    • If tool calls come from a country where you have no users or systems, check the source IP addresses in the Gateway Activity details.
    • Delegate the Governance Manager role to the team that owns the agent inventory, and the MCP Gateway Manager role to the team that runs the gateways.