Skip to main content
TrustGate is designed to be provider-agnostic. It sits in front of the AI surfaces you already use, and feeds the security tooling you already run.

​
LLM providers

Use TrustGate with any major provider — through native APIs or transparent proxy mode.
  • OpenAI (including Azure OpenAI), Anthropic, Google (Gemini / Vertex), Mistral, Cohere, AWS Bedrock.
  • Open-source and self-hosted models behind gateways that expose an OpenAI-compatible API.
  • Internal or fine-tuned models exposed via HTTP or gRPC.

​
AI platforms & frameworks

  • Agent frameworks that call LLMs and tools — LangChain, LlamaIndex, Semantic Kernel, AutoGen, CrewAI.
  • Retrieval stacks with vector stores, knowledge bases, and RAG pipelines.
  • Internal AI platforms that expose their own chat or completion APIs.

​
APIs & gateways

  • Put TrustGate behind an existing API gateway (for example Kong, Apigee, AWS API Gateway) to add AI-aware controls.
  • Or put TrustGate in front, for new AI-only routes where you want TrustGate to own routing, auth, and enforcement.

​
MCP servers & tools

TrustGate understands MCP natively, but at the Runtime Security layer — not via the AI Gateway. The AI Gateway carries the LLM request/response (which contains the tools[] definitions and tool_calls the model produces); the Runtime Security plugins inspect those payloads to:
  • Restrict which tools each route or identity may call.
  • Validate tool arguments against a schema.
  • Apply rate and budget limits per tool and per conversation.
MCP servers themselves are not routed through the AI Gateway; agents continue to talk to their MCP servers directly. See Agent & MCP security for the controls.

​
Observability & SIEM

  • Logs and events stream out as JSON over HTTP, file, or message bus.
  • OpenTelemetry traces and metrics are supported end-to-end.
  • Native SIEM connectors for Splunk, Elastic (ELK), IBM QRadar, Microsoft Sentinel, and Datadog — so AI detections can be correlated with identity, network, and application telemetry. See SIEM Integration.

​
Identity

  • SSO and SCIM for admin access.
  • Request-level identity is carried through as metadata on every event — so policies and logs can reason about users, apps, and tenants.