Analyze gateway traffic and token usage

The Fabric Gateway Analytics shows gateway traffic, errors, and LLM token usage across a selected window, with Overview, MCP, and LLM tabs. Each tab supports protocol filters (Overview — All protocols, MCP, and LLM) and time ranges (one hour, 24 hours, 7 days, and 30 days).

To access Analytics, go to Observability > Analytics in the left sidebar.

Overview

The overview surfaces the following metrics:

• Total requests, error rate, P95 latency, and active consumers.
• Traffic — Requests per minute by protocol.
• Errors — Client and server errors over time (4xx vs 5xx).
• P95 latency — 95th percentile request duration.
• Top routes — Routes with the most traffic in the selected window.
• Top consumers — Authenticated identities with the most requests.

MCP

MCP surfaces the following metrics:

• Total calls, error rate, P95 latency, and active consumers.
• Activity — Requests per minute by Tools, Resources, and Prompts.
• Top tools — Tools with the most calls in the selected window.
• Top consumers — Authenticated identities with the most calls in the selected window.
• Top routes — Routes with the most calls in the selected window.
• Top servers — Upstream MCP servers ranked by tool-call volume.
• Errors — Client and server errors over time (4xx vs 5xx).
• P95 latency — 95th percentile request duration.

LLM

LLM surfaces the following metrics:

• Total requests, total tokens, estimated cost, and error rate.
• Cost over time — Estimated spend per hour or model.
• Top models by estimated spend.
• Top consumers by estimated spend.
• Errors — Client and server errors over time (4xx vs 5xx).
• P95 latency — 95th percentile request duration.