> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ewake.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# On-Call Agent
> Automatically investigate alerts the moment they fire in Slack, before your engineer opens a terminal.
**What you'll get:** By the time your on-call engineer opens their laptop, ewake has already investigated the alert, probable cause ranked by confidence, supporting evidence assembled, next steps ready. The work that takes 30 minutes manually happens in seconds, automatically.
***
## What it does
The On-Call Agent monitors a configured Slack channel for alerts from your observability tools (Datadog, PagerDuty, etc.). When an alert fires, ewake immediately queries your connected sources, metrics, logs, recent deployments, past incidents, and posts a structured investigation response in the alert thread.
Your on-call engineer arrives to a Slack thread that already contains:
* The probable root cause, ranked by confidence
* Supporting evidence (relevant logs, metrics, recent commits)
* Suggested next steps
No manual log queries. No dashboard clicking. The context is already there.
***
## When to use it
* You have alerts routing to a Slack channel (`#alerts`, `#incidents`, `#monitoring`)
* Your team is on-call and needs faster first-response context
* You want to reduce the time between alert firing and understanding what's happening
***
## What you'll get
A typical ewake response in a Slack alert thread includes:
* **Summary**, one-line description of what happened
* **Probable cause**, top hypothesis with confidence indicator
* **Evidence**, log excerpts, metric spikes, recent deploys that correlate
* **Suggested next steps**, concrete actions to investigate or resolve
***
## Prerequisites
Before setting up the On-Call Agent, make sure these integrations are connected:
Datadog, Grafana, PagerDuty, or others
For the strongest alert investigation context, also connect:
Lets ewake reference past incidents and postmortems, surface patterns like "this looks like incident #47 from last month."
Sends deployment events to ewake so it can immediately correlate an alert with a recent deploy.
***
## How to set it up
In your ewake dashboard, navigate to the **Alert Response** tab.
Click **Create New Trigger**.
Enter a name that identifies this trigger (e.g. `datadog-alerts-prod`, `pagerduty-critical`).
Choose the channel where your alerts land, for example:
* `#alerts`
* `#incidents`
* `#monitoring`
Ewake must be present in this channel. [Add it here →](/get-started/connect-slack#add-ewake-to-a-channel)
Choose which bot ewake should listen to (e.g. Datadog, PagerDuty). This ensures ewake only responds to real alerts, not other messages in the channel.
Add an instruction prompt to guide how ewake responds. This is the main lever to customise tone and depth.
**Examples:**
```
Focus on identifying the root cause. Always include the 3 most recent deployments as context.
```
```
Be concise. Lead with the most likely cause. Include relevant log lines as evidence.
```
Click **Create**.
The On-Call Agent is active. The next time an alert fires in the selected channel, ewake will automatically respond in the thread.
***
Need help setting this up? Contact [support@ewake.ai](/support).