Meet the MCP Gateway

Take control while AI shifts from simple chatbots to autonomous agents that use the Model Context Protocol (MCP) to talk to data sources and tools. MCP Gateway sits inline between MCP clients (AI agents) and MCP servers (data/tool providers) to monitor, inspect, and protect this agent traffic.

┌──────────────────────────────────────────────────────────────┐
│                  NETSKOPE SECURITY FRAMEWORK                 │
├──────────────────────────────────────────────────────────────┤
│                                                              │
│  ┌───────────┐       ┌───────────────────────── ┐      ┌────────────┐
│  │MCP Client │       │Netskope AI Gateway Inline│      │MCP Servers │
│  │(Anthropic,│◄─────►│• Protocol Parsing        │◄────►│(Data/Tool  │
│  │ Cursor...)│       │• Malware/Injection Checks│      │ Providers) │
│  └───────────┘       │• Data Leakage Prevention │      └────────────┘
│                      └───────────────────────── ┘            |
│                                                              │
└──────────────────────────────────────────────────────────────┘

Live guardrails: Applies access controls, rate limits, and DLP rules to agent traffic in real-time.
Agent visibility: See exactly which tools your agents are triggering right inside your event logs.
Deep traffic inspection: Blocks malware, stops prompt injections, handles content moderation, and prevents data leakage.
Unified management: Manage MCP servers and permissions in your existing UI.

See MCP Gateway Overview for more information.

Policy Configurations for MCP Servers

Target specific agent capabilities with precision rules

We have integrated MCP Server controls straight into your primary policy builder so that you can manage agent permissions with precision. You can now build tailored rules and exclusions directly from Policies > AI Gateway.

The Old Way	The 1.5 Way
Security policies were limited to standard LLM endpoints. Agent tool calls were outside policy scope.	Policies can be configured directly for MCP Servers. Target exact capabilities (like “echo” or “get-resource-links”) and pair them with standard DLP Profiles.

For more information on configuring policies for MCP Servers, see AI Gateway Policies.

Quick Upgrades

Plug-and-Play Perplexity

Connect to Perplexity using your existing OpenAI SDK

We have added perplexity-openai as a new predefined AI provider using an OpenAI-compatible schema. If your teams are already sending traffic through the OpenAI SDK, they can route requests to Perplexity by updating the provider configuration—no code changes on the client side.

UX & Configuration Boosts

Cleaner Logs on SkopeIT

Instant traffic attribution on SkopeIT—no cross-referencing required

AI Gateway event logs on SkopeIT now display the explicit API token name, not just the token group. When an alert fires or you are reviewing activity, you immediately know which application or integration generated the traffic—without manually mapping token IDs.

See Monitoring AI Gateway Events for more information.

Rich Context Fields in AI Gateway Events

Goodbye look-alike string hashes, hello instant tracking context

Securing AI means knowing exactly what happened, who did it, how long it took, and how efficiently tokens were consumed. We have extended the telemetry in our event logs with three highly requested tracking fields.
You can instantly identify the exact application or user generating traffic and track precise API latency without tedious manual database lookups.

Category	Event Field	Technical Detail
Authentication	`x_ai_token_name`	Displays the explicit authentication token name used by the request.
Performance	`x_time_taken`	Measures processing duration from the first byte received to the last byte sent (in milliseconds).
Token Economy	`x_rs_ai_token_usage_cached_input`	Exposes cached input token usage of the AI request reported by the provider.

AI Gateway Release Notes Version - 1.5