WitnessAI Documentation
List
Start Here
Start Here
Analytics
Analytics
Conversations
Conversations
Users
Users
Alerts
Alerts
Policy
Policy
Administration
Administration
Apps & Resources
Apps & Resources
Release Notes
Release Notes
Administrator Guide
Administrator Guide
Model Context Protocol (MCP) Servers
The MCP Servers page provides visibility into Model Context Protocol (MCP) servers discovered across your environment. MCP servers extend AI applications and agents by providing access to external tools, data sources, and business systems.
Use this page to identify which MCP servers are being used, understand which tools they expose, determine which users and agents are interacting with them, and control user and agent access to MCP Servers and Tools through WitnessAI policies.
MCP Servers Overview
The MCP Servers page displays all MCP servers observed by WitnessAI during the selected time period. For each server, WitnessAI tracks:
- Server name and identifier
- Available tools
- Users accessing the server
- Agents utilizing the server
- Last observed activity
- Applied policy status
Administrators can use this information to identify approved integrations, monitor adoption, and restrict access to MCP servers that present security or compliance risks.
Filters
Use the filters at the top of the page to refine results:
Filter | Description |
Time Range | Limit results to a specific observation period. |
User | Display activity for a specific user. |
Department | Filter results by department. |
Group | Filter results by group membership. |
To find a server by name, type in the Search servers... box.
MCP Server Table
The table displays all discovered MCP servers matching the selected filters. Use the column picker in the table header to show or hide columns.
Column | Description |
MCP Server | Name of the MCP server and server identifier. |
Tools | Number of tools exposed by the MCP server. |
Users | Number of users who have accessed the server. |
Agents | Number of agents using the server. |
Last Seen | Most recent activity observed for the server. |
Policy | Current policy status applied to the server. |
Note: The server list refreshes on a schedule, about every five minutes by default, though the interval can differ by deployment. A newly seen server may not appear until the next refresh.
Policy Status
Each MCP server displays its current policy status.
Status | Description |
Allowed | The MCP server is permitted and available for use. |
Restricted | Access is partially limited based on policy controls. |
Blocked | Access to the MCP server is prohibited. |
Actions
Administrators can manage MCP server access directly from the table. The Actions column appears only for users who can edit policies.
Allow
Allows users and agents to access the MCP server according to configured policies.
Block
Prevents users and agents from accessing the MCP server.
Additional Actions
Each row's main button offers whichever of Allow or Block isn't in effect. The More actions menu has the other action, plus Remove from List when the server has an explicit rule. Removing it returns the server to the default.
Server Details
Click a server row to open Server Details. The header shows the server's First Seen and Last Seen dates, and the panel has three tabs: Tools, Users and Agents.
The Tools tab shows each tool's policy status as a read-only badge, so you can see how the server's policy lands on its tools. You set the policy for the server itself, from the table.
Use Cases
Common uses for the MCP Servers page include:
- Identifying newly introduced MCP servers
- Monitoring MCP server adoption across teams
- Reviewing tool exposure and usage
- Detecting unauthorized or unapproved integrations
- Restricting access to high-risk MCP servers
- Supporting governance and compliance requirements for AI agent ecosystems