NeuralTrust (TrustGate)

NeuralTrust builds security products for the agentic era — agent runtime security, agent gateway, agent posture management, and AI red teaming — and TrustGate is the gateway layer of that portfolio. Rather than letting each team secure its own AI traffic, TrustGate centralizes identity, routing, policy enforcement, and observability so that thousands of individual decisions collapse into one policy surface the security team controls. It covers three traffic classes in one gateway: model calls, MCP tool/connector access, and A2A (agent-to-agent) handoffs.

The gateway forwards end-user identity through every hop, supports per-agent and per-tool role-based access control, and produces cryptographic audit trails that survive internal security review and external audit. On the security side it performs inline prompt inspection, PII redaction through data masking, and semantic caching, aiming for under 100 ms added latency at p95 and throughput north of 20k requests per second per node in sustained enterprise load. It also handles practical operational concerns like AI cost visibility and AI API failover when a provider goes down.

Deployment is flexible to match compliance needs: a managed SaaS tier, a hybrid setup with a SaaS control plane and privately hosted data plane, and on-premises or air-gapped installs via Kubernetes and Helm charts for defense, government, and financial workloads. TrustGate is positioned as an open-source AI gateway with built-in security, with a free signup path and documentation and source available through GitHub, while enterprise governance, compliance, and self-hosted options are sold through sales conversations.