Monitor Token Consumption
Why this matters
One of the most powerful capabilities of AI Gateway is built-in token visibility. Because every LLM request passes through the proxy, the platform automatically tracks exactly how many tokens each client application consumed, which model was used, and when — all without any additional instrumentation.
This is what makes enterprise AI cost attribution possible. Instead of receiving a single monthly bill from your LLM provider with no breakdown, your platform team can see exactly which application consumed what — and charge back accordingly.
Note: This step is performed in the new MuleSoft UI.
Step 1 — Open the Monitoring tab
- In the left navigation, click LLM Proxies
- Click on your proxy
- Click the Monitoring tab
Step 2 — Review token consumption
The Monitoring tab breaks down token usage across two dimensions:
By model
See which model (gpt-5-mini, or any other route you configured) received traffic and how many tokens were consumed — prompt tokens and completion tokens shown separately.
By client application
Each application that requested access gets its own row. The my customer agent application you created shows the tokens consumed from the test request you sent in the previous step.
This is the value: An enterprise with 50 teams using different AI agents through the same gateway can see exactly who is consuming what — across every LLM provider — from one dashboard. No spreadsheets, no manual tracking, no surprises at billing time.
What you’ve seen
- Monitoring tab shows token consumption for your proxy
- Tokens are attributed to the
my customer agentclient application - Model breakdown shows which route handled the request
Continue to Summary & Next Steps →