Summary & Next Steps
What you did
| Step | What you did |
|---|---|
| Prerequisites | Confirmed access to the platform, verified the shared Omni Gateway, and set up Postman |
| Create LLM Proxy | Created an LLM Proxy with model-based routing to OpenAI GPT-4o Mini via the workshop endpoint |
| Verify Policies & Get Credentials | Confirmed the 6 auto-applied policies, navigated to Exchange, and obtained a Client ID and Secret |
| Test the Gateway | Sent a request through the gateway using Postman and received an LLM-generated response |
| Monitor Token Consumption | Viewed per-application and per-model token usage in the Monitoring tab |
Key takeaways
Developers authenticate with Client ID and Secret — not LLM keys. LLM provider API keys are stored encrypted in the platform. Developers never see them — this is the correct security posture for enterprise AI.
Model-based routing is transparent to the client. The client sends a model name in the request body. The gateway handles all provider-specific translation. Switching or adding providers requires no application code changes.
Token consumption is automatically tracked per client application. Every request through the gateway is attributed to the application that made it. At scale, this gives finance teams a complete, automatic breakdown of AI spend — by team, application, and model.
AI Gateway is built on Omni Gateway. All standard Omni Gateway capabilities apply — TLS, policy enforcement, logging. AI Gateway adds LLM-specific capabilities on top.
Resources
You’ve completed the MuleSoft AI Gateway workshop.