NetFoundry Adds Zero Trust MCP and LLM Gateways for AI Security

NetFoundry has expanded its AI Enclave solution with new zero trust Model Context Protocol (MCP) and Large Language Model (LLM) gateways, announced in a press release. The products aim to secure AI infrastructure by eliminating network exposure and providing identity-based access control.

The MCP Gateway allows access to MCP servers from any compatible client without exposing them to the network. It includes centralized management, role-based access control, multi-backend aggregation, and per-client session isolation. The LLM Gateway provides governed access to LLM providers such as OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Google Vertex AI, and private Ollama instances. It removes the need to distribute API keys and uses semantic routing to optimize requests by cost, latency, or data sensitivity.

Both gateways operate through NetFoundry’s Identity-First Reachability model, which assigns each AI agent, MCP server, and LLM endpoint its own cryptographic identity. Connections are outbound-only, encrypted, and continuously authenticated. This approach prevents unauthorized access and removes exposed attack surfaces.

NetFoundry also introduced an Accelerator Program offering early access to upcoming features, including an Agent2Agent network for identity-based agent communication. The new gateways and program are available for on-premise, hybrid, and cloud AI deployments.