MCP Gateway

Agent Identity Security MCP Gateway provides a controlled intermediary between AI agents and enterprise MCP servers, introducing a centralized control point for MCP communication with authentication, runtime access enforcement, response protection, and auditability.

Instead of connecting directly to a target MCP server, an AI agent connects through the MCP Gateway. The Gateway evaluates the interaction and forwards permitted requests to the configured target MCP server.

Core capabilities

MCP Gateway gives you five built-in controls across every agent-to-server interaction:

Table 1. MCP Gateway core capabilities
Capability What it provides
Bidirectional visibility Provides visibility into requests sent from agents to MCP servers and responses returned from MCP servers to agents.
Authentication Controls and validates access through the Gateway using the configured authentication method.
Access Enforcement Applies runtime access policies to control access to MCP servers and individual tools.
Response Protection Inspects supported MCP responses for sensitive information such as PII and secrets, and applies configured protection controls.
Auditability Records relevant MCP interactions and policy-enforcement outcomes in Event Logs for investigation and audit.

How the gateway works

When an agent sends a request, the gateway processes it in sequence:

  1. Connect through the Gateway The AI agent connects to the MCP Gateway instead of directly connecting to the target MCP server.
  2. Authenticate the connection The Gateway authenticates the incoming connection based on the configured authentication method.
  3. Evaluate runtime access Applicable runtime access policies determine whether access to the requested MCP server or tool is allowed, denied, or requires approval.
  4. Forward permitted requests Requests that are permitted are forwarded to the configured target MCP server.
  5. Protect the response Supported response-protection controls inspect responses for sensitive information such as PII and secrets before the response is returned to the agent.
  6. Record the interaction Relevant Gateway interactions and enforcement outcomes are captured in Event Logs for audit and investigation.

MCP Gateway provides the controlled communication and enforcement path, while Runtime Access Policies define what agents are permitted to access and invoke through that path.

Dashboard summary

The top of the MCP Gateway page shows two at-a-glance metrics:

Table 2. MCP Gateway dashboard metrics
Metric Description
Active Number of endpoints currently in Active status. Only active endpoints accept and proxy traffic.
Endpoints Total number of configured gateway endpoints across all statuses.

The toolbar provides two actions: Refresh reloads the list and stats, and Create Endpoint opens the endpoint creation form.

Endpoints inventory

The inventory is available in two views:

  • Cards view (default) Shows each endpoint as a tile with its name, description, status badge, authentication type, request count, target URL, Gateway URL, and inline action buttons.
  • Table view Shows the same information in a sortable grid. Select any column header to sort.
Note: The Gateway URL is the address your AI agents and MCP clients must be configured to use. Provide this URL in place of the target server URL in your agent's MCP server configuration. Agent Identity Security authenticates, sanitizes, and forwards every request transparently.
Table 3. Endpoint inventory columns
Column Description
Proxy Name Human-readable name for the endpoint. Supports text filtering: Contains, Equals, Not equals, Does not contain. Long names truncate with a tooltip.
Gateway URL The public proxy URL agents use to send MCP requests. Format: https://<host>/mcp-proxy/v1/<endpoint-id>. Select the copy icon next to the URL to copy it to your clipboard.
Target URL The actual MCP server URL the gateway forwards requests to.
Auth The authentication method used when forwarding requests to the target server: None, Bearer, API Key, JIT Entra Token, mTLS, or OAuth 2.0.
Requests Cumulative number of MCP requests proxied through this endpoint since it was created. Supports number-range filtering.
Last Request Timestamp of the most recent request. Shows Never for endpoints that have not yet received any traffic. Supports date-range filtering.
Status Active Endpoint accepts and proxies agent requests. Disabled Endpoint rejects all incoming requests; no traffic is forwarded.
Actions Inline buttons per row: Edit, Deactivate or Activate (based on current status), Copy URL, and Delete.