Skip to main content
fal exposes a Prometheus-compatible metrics endpoint that you can scrape with any monitoring tool. Use it to build custom dashboards, set up alerts on queue depth or error rates, and feed fal metrics into the same observability stack you use for the rest of your infrastructure. The endpoint returns metrics in Prometheus exposition format, so it works with Grafana, Datadog, New Relic, Splunk, or any tool that can scrape a Prometheus target. Responses are cached for 10 seconds, so set your scrape interval accordingly.

Endpoint

curl -H "Authorization: Key $FAL_KEY" \
  https://api.fal.ai/v1/serverless/metrics

Available Metrics

MetricLabelsDescription
fal_app_runnersapplication, state, machine_typeNumber of runners currently allocated
fal_app_queue_sizeapplicationRequests waiting in queue
fal_app_concurrent_requestsapplicationRequests being actively processed
fal_app_requests_completedapplication, method, statusRequests completed in the last minute
fal_app_requests_receivedapplication, methodRequests received in the last minute
fal_app_request_latencyapplication, leCompleted requests bucketed by latency
The fal_app_runners metric tracks runners across three states: idle (warm, waiting), running (processing), and pending (cold start in progress).

Integration

Add the endpoint as a Prometheus data source in your monitoring tool. The only requirement is passing your API key in the Authorization: Key ... header. Set the scrape interval to at least 10 seconds since responses are cached at that frequency.

Example PromQL Queries

# Active runners by app
fal_app_runners{state="running"}

# Queue depth across all apps
sum(fal_app_queue_size) by (application)

# Request rate (completed per minute)
fal_app_requests_completed

# P99 latency estimation
histogram_quantile(0.99, fal_app_request_latency)
All metrics are gauges. The fal_app_request_latency metric uses histogram-style buckets (labeled by le) for latency distribution analysis.

Platform API Reference

Full API specification for the metrics endpoint