Hyphen

Application Health Checks

Learn how to use Hyphen Agent to run one-time application health checks or configure continuous monitoring using logs and HTTP request metrics.

Ask Hyphen Agent to check one deployed app in one environment. For example, ask “Check the health of Checkout API in Production now.” You need access to Agent and permission to manage the app.

Table of Contents

Run a check

Ask “Analyze the logs for Checkout API in Production” to run a one-time application health check.

Agent checks the available logs and HTTP request metrics and investigates signs of a problem. The result appears in the same chat, even if Agent finds no problem. If a source is unavailable, the result explains which evidence Agent could inspect. Active checks and investigations use agent minutes.

You can also start an ApplicationHealthCheck through MCP. Read the result in the task conversation with get_agent_task.

A manual check runs once. You can run it without on-call coverage or a continuous monitor. In environments with on-call coverage, a confirmed problem can create an incident and page responders.

Continuous monitoring

To keep watching the app, enable monitoring under Project Settings → On-call → Agent App Monitoring. Choose an app in a covered environment and set the interval. You can also ask Agent to watch an app continuously from chat.

Manual checks use the same telemetry as continuous monitoring. Connect an available cloud log or metrics source, or configure an APM connection for the app and environment.

Telemetry Sources

Agent reads HTTP request health and logs from these sources for the selected app and environment:

Source HTTP request health Logs
Google Cloud Run Cloud Run request metrics. Container logs.
AWS ECS Application Load Balancer target group metrics: request count, target 5xx responses, and p95 response time. Container logs.
Azure Container Apps Azure Monitor Requests metric by HTTP status category. Console logs from the Log Analytics workspace of the Container Apps environment.
Cloudflare static sites Request and failure counts for the current deployment. Optional Worker logs.
Cloudflare containers Unavailable from Cloudflare. Optional Worker and container logs.
Datadog APM Trace metrics of the inbound HTTP operation of the service. App logs.
New Relic APM APM request metrics. App logs.
SigNoz APM APM request metrics. App logs.

Some sources have limits:

  • AWS metrics include only responses from the app target group. Failures at the load balancer or ingress are not included. Agent reads complete minutes only, and CloudWatch can lag by a few minutes.
  • Azure request metrics can arrive late. Agent cannot verify that Azure sent all request counts for the check window.
  • Azure logs need a Container Apps environment that sends logs to Log Analytics. Hyphen also needs query access to that workspace.

When a source covers only part of the app, the result tells you that the request health is incomplete.

Historical Log Analysis runs

Open Agent → Tasks and filter by Log analysis to read saved findings and GitHub issue actions from past runs. Log Analysis schedules are retired. To check the app again, request an application health check.