How to Read the MCP Gateway Section
Use this topic when you want to understand what the MCP Gateway section of the AI Governance Dashboard shows and how to use it to monitor agent activity, tool usage, and data protection across your gateway connections.
The MCP Gateway dashboard gives you a centralised view of everything happening through your gateway connections. You can see which agents and users are active, which tools are being called, and whether any DLP violations have been detected.
There are two places where MCP Gateway data appears:
- The AI Tool Landscape tab on the AI Governance Dashboard — gives a high-level summary of all gateway connections alongside your sanctioned AI apps
- The individual connector view — a detailed drill-down for each gateway connector showing activity charts, exposed tools, and recent calls.Note: The AI Tool Landscape tab also contains a Sanctioned AI Apps section, an Unsanctioned AI Apps section, and an Identity Landscape section. This topic covers only the MCP Gateway Connections section. See Read the AI Governance Dashboard for others.
The AI Tool Landscape tab — MCP Gateway Connections section
Go to AI Security and then click on Dashboard and click the AI Tool Landscape tab.

Scroll down past the Sanctioned AI Apps section to find MCP Gateway Connections.

Each connected gateway connector appears as a card showing:
| Metric | What it tells you |
|---|---|
| Active Users | The number of users who have made tool calls through this connector in the selected time period |
| Agents | The number of registered agents actively using this connector — shown in purple to highlight agent activity separately from user activity |
| Tools | The total number of MCP tools exposed by this connector |
| Calls | The total number of tool calls made through this connector in the selected time period |
| DLP Violations | The number of tool calls that triggered a DLP policy — shown in red when violations are present Note: In this Early Access (EA) release, this feature is not
available.
|
Click on any connector card to open the detailed connector view.
The individual connector view — Atlassian example
Clicking a gateway connector card opens a detailed view for that connector. The top of the screen shows a summary strip with key metrics for the selected time period.

Summary strip
| Metric | What it tells you |
|---|---|
| Active Users | Total number of individual users who made tool calls through this connector |
| Agents | Total number of agents active on this connector, shown with a note of how many platforms they span |
| Tools Exposed | Total tools available through this connector — the number in smaller text below shows how many of those tools are marked as sensitive |
| Calls | Total tool calls in the selected period |
| DLP Violations | Total DLP violations detected — the number below shows how many were blocked rather than just logged or warned Note: In this Early Access (EA) release, this feature
is not available.
|
Top Active Users
This table shows the users who made the most tool calls through this connector in the selected period, along with their department and when they were last seen.

Use this to identify users with unusually high call volumes or users accessing sensitive tools who you would not expect to be active on this connector.
Agents Connected

This table sits alongside Top Active Users and shows the agents currently active on this connector. For each agent it shows:
- The platform the agent is running on (for example, Copilot Studio, AWS Bedrock, or Internal)
- The total number of calls the agent has made
- The agent's risk level — Low, Medium, or High — based on the type of data it is accessing and the DLP violations associated with it
Use this to identify high-risk agents that may need stricter access control policies or scope restrictions.
MCP Tools / Resources Exposed

This table shows the top tools being called through this connector, sorted by call volume.
For each tool it shows:
| Column | What it tells you |
|---|---|
| Tool | The MCP tool name as exposed by the connector |
| Description | A plain-language description of what the tool does Note: In this Early Access (EA) release, this feature is not available.
|
| Calls | How many times this tool has been called in the selected period |
| Sensitivity | Whether the tool accesses sensitive data — shown as Sensitive in orange when applicable Note: In this Early Access (EA) release, this feature
is not available.
|
Use this table to understand which tools are most heavily used and whether sensitive tools are being accessed at a volume that warrants tighter policy controls.
Recent Calls
This table shows the most recent individual tool calls made through this connector.

For each call it shows:
| Column | What it tells you |
|---|---|
| Time | When the tool call was made |
| Actor | The agent or user that made the call — agents are shown with an agent icon, users with a person icon |
| Method | The MCP tool call that triggered the violation |
| Classification | The data classification of the content involved — for example, Customer PII or Financial Note: In this Early Access (EA) release, this feature is not
available.
|
| Status | Whether the call was Allowed, Warned, or Blocked Note: In this Early Access (EA) release, this feature is not available.
|
Use Recent Calls to investigate specific activity — for example, if ARIA flags an unusual pattern, you can trace the exact tool calls here.
DLP Violations

This table shows tool calls that triggered a DLP policy, ordered by most recent.
For each violation it shows:
| Column | What it tells you |
|---|---|
| Time | When the violation occurred |
| Actor | The agent or user responsible |
| Method | The MCP tool call that triggered the violation |
| Classification | The type of sensitive data detected |
| Severity | How serious the violation is — Critical, High, or Medium |
| Action | What the gateway did — Blocked, Warned, or Logged |
Use DLP Violations to investigate incidents, identify repeat offenders, and decide whether existing policies need to be tightened. A Critical severity violation that was only Logged — rather than Blocked — may indicate a policy that needs to be updated to Deny.