Datadog Monitor Digest
Daily digest of triggered Datadog monitors, delivered to your InTouch subscribers.
Use case
Datadog fires monitors around the clock, but not every alert warrants a page. This workflow rolls up the monitors that are currently in a non-OK state into a single, once-a-day summary so on-call engineers and team leads can review what tripped over the last 24 hours without logging into the Datadog console.
What it does each run
- fetch-monitors (
datadogtool) — calls the Datadoglist-monitorsoperation, filtered to monitors currently in anAlert,Warn, orNo Datastate. Like every InTouch connector, it publishes exactly two outputs:{{fetch-monitors.result}}(the raw JSON API response as a string) and{{fetch-monitors.status}}(the HTTP status code). - summarize (
aitool) — feeds the raw{{fetch-monitors.result}}JSON to the model and produces a concise human-readable markdown digest ({{summarize.answer}}): a headline count followed by one bullet per triggered monitor, ordered Alert → Warn → No Data. - notify-digest (
messagetool) — wraps{{summarize.answer}}(plus the API status from{{fetch-monitors.status}}) into a single notification and sends it to the workflow's contacts.
If the Datadog API call fails, alertOnError: true notifies the intouch user.
Setup
Credentials
apiKey(datadog) — thefetch-monitorstask'sapiKeyproperty is set to the placeholder<set via credential>. Supply your real Datadog API key by attaching a credential (or runtime env) to the task; never hardcode the key in the YAML. Generate the key under Organization Settings → API Keys in Datadog. The credential only needs read access to monitors.- AI provider — the
summarizetask uses the built-inaitool withclaude-sonnet-4-6. It uses whichever AI provider/credential is configured for the InTouch assistant; no per-task API key is needed in the YAML.
Subscribers
Add at least one contact to this workflow so the digest has somewhere to land
(email, Slack, etc.). The intouch user is wired up for error alerts by
default.
Scheduling
Attach a daily schedule (for example, every morning at 08:00) so the digest arrives at the start of each business day. Run on demand at any time to get an immediate snapshot of currently-triggered monitors.