Monitor API usage
What this achieves
Section titled “What this achieves”API usage reports what your integrations are actually doing: how many requests they make, how fast they come back, and how many fail. It is the page that tells you an integration has quietly stopped, which is otherwise noticed by whoever was waiting for the data.
Required role: Administrator.
- Go to Integrations → API Usage.
- Read the four headline metrics.
- Read Request Volume (Last 30 Days) for the shape of the traffic.
- Check Top Endpoints and Response Status Codes.
- Open a key from the API Keys list to see its own request log.
- Filter that log by method, endpoint, status code, or date range.
The headline metrics
Section titled “The headline metrics”| Option | Description |
|---|---|
| Total Requests (30d) | Requests made in the last thirty days. |
| Avg Latency | Average response time, in milliseconds. |
| Error Rate | The share of requests that failed. |
| Active Keys | How many keys are currently live. |
Response status codes
Section titled “Response status codes”| Option | Description |
|---|---|
| 2xx Success | The request worked. |
| 3xx Redirect | The request was redirected. |
| 4xx Client Error | The request was wrong — a bad path, a missing scope, a rate limit, or a malformed body. |
| 5xx Server Error | The request failed on Humavera’s side. |
The split between 4xx and 5xx is the first diagnostic. A wall of 4xx is your integration asking for something it should not be asking for, most often a scope the key does not carry or a rate limit it keeps hitting. A wall of 5xx is a different conversation.
Example: HC Corp’s reporting job showing a steady 4xx rate against one endpoint is almost always a missing scope on the key. Nothing about the key looks wrong in the list; the request log is where it shows.
The per-key request log
Section titled “The per-key request log”Opening a key shows its individual requests, with the Method, Endpoint, Status, Latency, Time, and originating IP.
Filter it by method, by endpoint text, by status code, or by date range. Filtering to 4xx over the last week and reading which endpoints appear is the fastest way to find what a failing integration is trying to do.
The originating IP is worth a glance. Requests from somewhere you do not recognise, on a key you thought only one system held, is the signal to revoke that key.
What is not counted here
Section titled “What is not counted here”The page states it directly: usage is logged for requests authenticated with API keys. Traffic from people using Humavera in a browser is not counted on this page.
So a low figure here does not mean the product is idle. It means your integrations are idle, which is exactly the question this page answers.
An empty page is a finding
Section titled “An empty page is a finding”Last used: Never against a key, or no traffic at all where you expect some, is not a display problem. Either the integration was never finished, or it is authenticating some other way, or it stopped.
Example: HC Corp’s service desk integration showing no requests for eleven days does not raise an alert anywhere. It shows up here as a flat line, and the person waiting for new-joiner tickets to appear finds out later.
What happens next
Section titled “What happens next”Use the request log to fix what is failing — a missing scope means editing the key’s scopes, and repeated rate-limit rejections mean raising the key’s limit or slowing the caller. Where a key shows traffic you cannot account for, revoke it and issue a replacement.
Related
Section titled “Related”© 2025-2026 Humavera Documentation - BPilot Ltd. All Rights Reserved