Core Architecture
AI Agent API Gateway
The TrustGate AI Agent Gateway acts as the single choke-point for all autonomous agent traffic in your organization. It intercepts, sanitizes, and routes agentic requests to underlying models (OpenAI, Anthropic, or local open-source).
tg_sk_...
Your App
Standard OpenAI SDK
TrustGate Proxy
OpenAI
Anthropic
Gemini
Mistral
Universal Routing
TrustGate uses LiteLLM under the hood to normalize requests. You don't need to change endpoints when switching providers. Just change the model parameter in your request.
POST /v1/chat/completions
{
"model": "claude-3-opus", // Or "gpt-4", "gemini-pro"
"messages": [...],
"stream": true
}
"model": "claude-3-opus", // Or "gpt-4", "gemini-pro"
"messages": [...],
"stream": true
}
The Provider Key Vault
Your clients never touch raw provider keys (e.g., sk-proj-...). These are stored securely in the TrustGate Vault (AES-256 encrypted).
- Client uses only Agent Key (tg_sk_...)
- Provider keys injected Server-Side
- Rotation happens in Dashboard, not Code
- Prevents Key Leaks in Frontend Apps