Anomaly Detection
Scout detects unusual performance behavior automatically. No thresholds to configure, no baselines to set. Your app's patterns are the baseline.
How it works
Detection that learns your app's behavior
Most alerting tools ask you to pick a number. "Alert when response time exceeds 500ms." But what if 500ms is normal for your reporting endpoint and terrible for your login page? Threshold-based alerting treats every endpoint the same.
Scout compares each endpoint against its own recent behavior, checking every minute. An endpoint that normally responds in 50ms spiking to 200ms triggers an alert. An endpoint that always takes 400ms running at 410ms does not. The baseline is the app, not your guess.
Per-endpoint baselines
Every endpoint and background job has its own baseline. Detection adapts to each one's normal variability, not a global threshold.
Instant history on setup
When you enable detection, Scout backfills against the past week of history so you see what would have been caught before the next incident happens.
Tunable sensitivity
Three presets (high, medium, low) plus a live preview slider that shows what would have been detected over the past week. Nothing saves until you confirm.
Seasonality
Recurring Patterns Are Handled Automatically
Most apps have natural performance rhythms. Scout cross-checks every spike against the same time 24 hours ago and the same time one week ago. Only deviations that are unusual for that time of day and day of the week get flagged. Nothing to configure.
Daily patterns
Nightly cron jobs, end-of-day batch processing, and daily traffic peaks are recognized and suppressed automatically.
Weekly patterns
Monday morning spikes, weekend lulls, and weekly billing runs are treated as normal for that day of the week.
Live preview when tuning
The sensitivity slider updates chart markers in real time so you can see exactly how many past events each level would have caught before saving.
Alerts that respect your attention
Anomaly notifications fire once when the event opens. No re-notifications while the anomaly is active. No alert storms during an incident. When an endpoint returns to normal, the event closes quietly. Notifications go through Scout's existing channels: Slack, PagerDuty, email, and Discord.
Included in every plan
Anomaly detection is available on all Scout plans, including the free tier. No upgrade required. Works with Ruby, Python, Node.js, PHP, and Elixir applications and any background job framework Scout instruments.
AI Native
Query anomalies from your AI coding agent
The Scout MCP server includes tools for listing and inspecting anomaly events. Your AI coding agent can ask which endpoints are behaving abnormally, see deviation magnitude, and correlate anomalies with recent deploys or code changes.
The Scout CLI provides the same access via scout anomalies. Filter by state, metric type, endpoint, or date range. The public REST API is available for custom integrations.
Pairs well with: Alerting for hard SLA thresholds, Scout MCP server for pulling anomaly data into AI workflows, and Scout CLI for querying anomalies from the terminal.
Add Scout to Your App
With monitoring that "just works" for Rails, Django and more, get running in minutes with Scout's streamlined setup. No extra code needed - just deploy our agent and start getting insights in 5 minutes.
Get started in under 3 minutes with our easy to install Ruby agent.
Install our gem. In your Gemfile:
gem 'scout_apm'Shell:
bundle installDownload your customized config file, placing it at config/scout_apm.yml.
Deploy.
Get started in under 3 minutes with our easy to install Python agent.
Install the package:
pip install scout-apmConfigure in your settings.py (Django):
# settings.py
INSTALLED_APPS = [
"scout_apm.django", # should be listed first
# ... other apps ...
]
# Scout settings
SCOUT_MONITOR = True
SCOUT_KEY = "[AVAILABLE IN THE SCOUT UI]"
SCOUT_NAME = "A FRIENDLY NAME FOR YOUR APP"Alternatively, use environment variables: SCOUT_MONITOR, SCOUT_NAME, SCOUT_KEY.
Deploy. Data appears in the Scout UI within approximately 5 minutes.
Get started in under 3 minutes with our easy to install PHP agent.
Install the package:
composer require scoutapp/scout-apm-laravelOptionally install the PHP extension:
sudo pecl install scoutapmConfigure in your .env file and deploy.
Get started in under 3 minutes with our easy to install Elixir agent.
Add to your mix.exs and run mix deps.get:
{:scout_apm, "~> 2.0"}Download your customized config and place it at config/scout_apm.exs.
Add instrumentation to lib/your_app_web.ex:
defmodule YourApp.Web do
def controller do
quote do
use Phoenix.Controller
use ScoutApm.InstrumentationDeploy.
FAQ
Anomaly Detection FAQ
What is anomaly detection in Scout?
Anomaly detection automatically identifies unusual performance behavior in your endpoints and background jobs without manual threshold configuration. It compares current response times against recent patterns, accounts for daily and weekly seasonality, and alerts only when deviations are meaningful. It checks every minute and is available for all Scout-supported languages: Ruby, Python, Node.js, PHP, and Elixir.
How is anomaly detection different from threshold alerting?
Threshold-based alerting fires when a metric crosses a fixed number you set (e.g., response time above 500ms). Anomaly detection fires when a metric deviates meaningfully from that endpoint's own recent behavior. An endpoint that normally responds in 50ms spiking to 200ms is worth investigating, even though 200ms would be fine for a different endpoint. Scout checks two seasonal windows (24 hours ago and 7 days ago) to suppress alerts for recurring patterns like nightly jobs or Monday traffic spikes.
Does anomaly detection require configuration?
No. Enable it on an endpoint and Scout immediately analyzes the past week of data. Detection starts within 30 minutes of live traffic. You can optionally tune sensitivity and set a minimum duration before an event fires, but the defaults work well for most endpoints.
Can I query anomaly events from Claude Code, Cursor, or other AI coding agents?
Yes. The Scout MCP server includes tools for listing and inspecting anomaly events. Your AI coding agent can query which endpoints are behaving abnormally and see deviation details. The Scout CLI provides the same access via scout anomalies. There is also a public REST API for custom integrations.
What metrics does anomaly detection cover?
Average response time and 95th percentile response time for endpoints and background jobs. It does not currently cover throughput or error rate — use threshold alerting for those.
Does it work with background jobs?
Yes. Enable anomaly detection per background job the same way you enable it per endpoint. It works with Sidekiq, Celery, Resque, Oban, Solid Queue, and any other background job framework Scout instruments.
Will it fire every time a busy period kicks in?
No. Scout cross-checks spikes against the same time 24 hours ago and the same time one week ago. Recurring patterns are suppressed automatically. Only deviations that are unusual for that time of day and day of the week are flagged.
Is anomaly detection included in all Scout plans?
Yes. Anomaly detection is available on all Scout plans, including the free tier. No upgrade required.