> ## Documentation Index
> Fetch the complete documentation index at: https://docs.whawit.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# On-call features
> Everything WHAWIT's on-call does — agentic investigation, schedules and rotations, escalation policies, escalation teams, multi-channel paging, guided setup, and the live activity view
## At a glance
| Area | What you get |
| ------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------- |
| **Agentic investigation** | Five specialist AI agents work the incident before anyone is paged. No runbooks to write. |
| **Schedules** | Timezone-aware rotations built from layers, with restrictions, handoff times and overrides. |
| **Escalation policies** | Ordered levels with per-level acknowledgement timeouts, three kinds of target, and repeat. |
| **Escalation teams** | Named groups a policy can page as a unit. |
| **Paging** | WhatsApp, SMS, Slack, Microsoft Teams, Discord, email — per escalation level. |
| **Guided setup** | Describe your team and WHAWIT proposes the whole configuration. |
| **Agent cards** | Every on-call agent in the project on one screen — status, objective, escalation wiring, provider and cadence, with start/pause in place. |
| **Live activity** | Watch the agents work, in a diagram of the team and its tools. |
| **Knowledge base** | What each incident established, written back for the next one. |
| **Incident coordination** | A dedicated Slack or Teams channel, roles, timeline, drafted postmortem. |
| **Migration** | Ingest alerts from PagerDuty and Opsgenie while you move across. |
***
## Agentic investigation
The part that is not scheduling. A standing team of five AI agents —
**Incident Commander**, **Log Analyst**, **Infrastructure Specialist**,
**Communications Lead**, **Knowledge Engineer** — works each incident, deciding
what to look at next from what it just found.
* **No rules to pre-write.** No runbook, no alert-to-action mapping, no decision
tree to keep current.
* **Runs on a cycle**, triggers immediately on an alert, and schedules its own
follow-ups. The cycle tightens while an incident is live.
* **Nine tools** across the team: query logs, group errors, search your
repository, search the knowledge base, analyze context, open and update
incidents, notify responders, attempt recovery.
* **The page carries the findings** — root-cause hypothesis, evidence, suggested
fix.
Full detail in [Agentic on-call](/on-call/agent-teams).
## Schedules and rotations
* **Layers.** A schedule is built from independently rotating layers — Primary,
Secondary, as many as you need.
* **Rotation types.** Daily, weekly, or custom with a shift length in hours.
Custom is how follow-the-sun coverage gets built.
* **Timezone per schedule.** An IANA timezone on the schedule itself, so
handoffs land at the right local hour.
* **Handoff times.** An explicit local time of day that a shift changes hands.
* **Restrictions.** Limit a layer to particular days and hours — business hours
only, weekends only, weeknights only. Windows may cross midnight.
* **Overrides.** Substitute a person into any window for leave or a swap, scoped
to one layer or to the whole schedule, with a note explaining why.
* **Rotation order.** Members rotate in the order listed, so the handoff
sequence is explicit rather than inferred.
Detail in [Schedules and rotations](/on-call/schedules).
## Escalation policies
* **Ordered levels.** Each names its targets and how long to wait.
* **Per-level acknowledgement timeout**, in minutes. No acknowledgement, next
level, automatically.
* **Three target kinds** — an individual, an entire on-call schedule (resolved
at escalation time, so it always reaches whoever is genuinely on call), or an
escalation team.
* **Per-level channels.** Start on Slack during the day, end on SMS at 3 AM.
* **Repeat.** Re-run the whole chain a set number of times if the incident is
still unacknowledged.
* **Organization default.** A fallback policy so nothing is ever raised with no
escalation path.
* **Policy resolution by specificity** — the on-call agent's policy, then the
client's, then the organization default. Failing all three, a chain is derived
from the agent's rotations. The [agent's card](/on-call/agents) shows how each
rotation resolves.
* **Preview.** Dry-run a policy and see exactly who it would page at each level,
right now or at any timestamp you choose, with current rotations and overrides
applied.
Detail in [Escalation policies](/on-call/escalation-policies).
Preview is the feature most worth knowing about. Point it at 03:00 on a Sunday
and it will tell you whether your escalation quietly resolves to nobody —
which is the failure you do not want to discover during an outage.
## Escalation teams
A named group with its own members, targetable by an escalation level as a
single unit. Useful when a whole team should be pulled in at once rather than
one person at a time, and it keeps policies stable as team membership changes.
## Notifications and paging
Six channels, configurable per escalation level:
Every page carries the severity, the summary, what the investigation found, and
a one-tap acknowledge link that stops the escalation immediately.
WhatsApp and SMS deliver to the phone number on the user's profile. A user
without one is skipped for those channels while the escalation continues on
the others — so it fails quietly. Fill phone numbers in before relying on an
SMS level.
Detail in [Notifications](/on-call/notifications).
## Guided setup
You do not have to hand-build the first configuration. WHAWIT will propose a
complete one — escalation teams, schedules and policies — from a short
description of your team. Three ways to give it the roster:
| Mode | How it works |
| ------------------------- | ------------------------------------------------------------------------ |
| **From your directory** | Uses the people already in your organization. |
| **From a list of emails** | Paste addresses. Anyone who is not yet a member is flagged to invite. |
| **Describe it** | Write what you want in plain language and let WHAWIT work out the shape. |
Then a handful of questions — timezone, rotation type, handoff time, whether you
want a single tier, two tiers, or follow-the-sun, the acknowledgement timeout,
and how many times to repeat — and it returns a proposal you review and adjust
before anything is created.
Detail in [Guided setup](/on-call/setup).
## Live activity
The agent team is not a black box. Clicking an agent card that carries the
**AI Agents** badge opens a view of the team as a diagram — the agents, the
tools they are reaching for, the knowledge base they are reading and writing,
and where each investigation ended up — plus an activity timeline of what
happened on an incident and when. The [agents screen](/on-call/agents) is the
front door to it.
## Knowledge base
The Knowledge Engineer writes back what each incident established. Recurring
failures are met with what the previous occurrence cost you to learn, instead of
being investigated from scratch every time.
This is the difference between automation that decays and automation that
compounds. A runbook records what somebody knew when they wrote it; a
knowledge base the agents both read and write records what the system has
actually done since.
## Incident coordination
Paging gets the right person awake. Coordination is what happens next:
* A dedicated **Slack or Microsoft Teams channel** for the incident, with
responder roles assigned.
* A **living timeline** of what was done and when.
* **Drafted status updates**, so responders respond instead of narrating.
* A **drafted postmortem** built from the timeline once it is over.
* **Tickets synced** to Jira, GitHub or Azure DevOps.
## Alert intake and migration
WHAWIT correlates and deduplicates the alerts arriving from your connected
providers, so one underlying failure raises one incident rather than forty
notifications.
It can also **ingest alerts from PagerDuty and Opsgenie**, which is how teams
evaluate it against live incidents before moving any rotation. A migration path,
not a dependency — see [Migrating](/on-call/migrating).
## Related
Get a first configuration standing.
Direct answers about coverage, migration and what is included.