> ## Documentation Index
> Fetch the complete documentation index at: https://docs.devinenterprise.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Devin Oncall

> Set up Devin Oncall to investigate, fix, and mitigate alerts, bugs, and incidents, and act on insights from your oncall.

Devin Oncall investigates, fixes, and mitigates your alerts, bugs, and incidents, and gives you insights from your oncall to act on. Devin starts as soon as an alert, bug report, or incident fires, and uses your telemetry, code, deployments, runbooks, and past incidents to find the root cause, remediate the issue, and open a fix PR.

Oncall has three products:

* **[Responders](#responders)**: investigate and fix alerts, bug reports, and support questions as they arrive.
* **[Incidents](#incidents)**: find the root cause, mitigate, and drive each incident to resolution.
* **[Reports](#reports)**: act on insights into what is driving your alerts, bugs, and incidents, such as alert thresholds to tune and recurring classes of issues to fix.

## Setup

Oncall needs a one-time setup per organization. During setup, Devin learns your systems from your telemetry tools and code, and writes skills that every Oncall session uses.

1. Open **Oncall** in the sidebar and click **Begin setup**.
2. Connect at least one telemetry tool. You can skip this step if a telemetry tool is already available in Devin's environment.
3. Click **Finish**. Devin starts learning your systems, which usually takes 1-2 hours. You are notified when Oncall is ready.

Responders, Incidents, and Reports become available once setup is complete.

<Note>Setup requires permission to manage Oncall. Only organization admins can connect tools for the whole organization.</Note>

### Oncall settings

The settings gear in the Oncall header opens what every Oncall session shares:

* **Instructions**: guidance every Oncall session follows, written in markdown. Think of it as an `AGENTS.md` just for Oncall.
* **Skills**: the skills Devin generated about your systems during setup. Set **Skill update frequency** to **Weekly** or **Monthly** to keep them current, or click **Update skills now**.

## Responders

A responder watches a source of alerts or reports and has Devin investigate each new event. Devin groups repeated events into a single issue, so a noisy alert is investigated once.

Use responders for:

* Production alerts posted to a Slack or Microsoft Teams channel by your monitoring tools.
* PagerDuty incidents, where Devin posts its investigation as incident notes.
* Bug reports and support questions in a channel like `#eng-bugs`.

### Create a responder

1. On the **Oncall** page, click **Create responder**.
2. Enter a **Name**, and choose the **Responder type**. The type can't be changed after creation.
3. Add one or more **Triggers**.
4. Optionally adjust the session settings, connectors, and runbook.
5. Click **Create responder**.

<Note>A runbook is optional. Devin Oncall already knows how to investigate alerts and bugs, so add one only for steps specific to your team.</Note>

### Responder types

Alerts and bug reports need different handling, so each responder type runs a flow built for it. The type also sets how the responder's issues are tracked on its board.

| Type | Use it for | What Devin focuses on |
| - | - | - |
| **Alerts** | Alerts on production issues | Assessing impact and severity quickly, then finding the root cause and a mitigation |
| **Support + Bugs** | Bug reports and support questions | Reproducing the issue, finding the root cause, and fixing it in code |

### Triggers

A responder watches one integration at a time and can have up to 10 triggers. Devin responds when any of them match.

| Source | Triggers |
| - | - |
| **[Slack](/integrations/slack)** | A message in a channel, or a user group mention (for example `@oncall`) |
| **[Microsoft Teams](/integrations/microsoft-teams)** | A message in a channel |
| **[PagerDuty](/integrations/pagerduty)** | Incident triggered, acknowledged, resolved, or updated |

Add conditions to a trigger to narrow which events Devin responds to.

### Responder settings

| Setting | What it does |
| - | - |
| **Agent mode** | The agent mode for sessions this responder starts |
| **Run as** | The identity sessions run as |
| **Connectors** | The MCP connectors this responder can use |
| **Runbook** | Optional. Devin Oncall already knows how to investigate alerts and bugs. Add team-specific guidance, such as how you assign priority or where to file issues |
| **Limits** | ACU and invocation limits, as in [automations](/product-guides/automations#limits-and-safeguards) |
| **Advanced** | Network policy, and metadata key-value pairs used to organize responders and match them to reports |

## Incidents

Incidents is built for teams that create a new Slack channel for each incident. Devin joins the channel, finds a likely root cause, recommends mitigations, and keeps investigating as the incident develops. It posts updates, answers questions in threads, and maintains an incident report.

### Enable incidents

1. On the **Oncall** page, open the **Incidents** tab and click **Enable incidents**.
2. Turn on **Enable incidents**.
3. Set the **Slack channel prefix** your incident channels use, for example `incident-`.
4. Choose a **Response mode**, then save.

Devin starts an investigation when it joins a Slack channel whose name starts with the prefix. Invite Devin to the channel, or turn on **Auto-join incident channels**.

<Note>Incidents require [Slack](/integrations/slack) to be connected for your organization. In an enterprise, one organization at a time can have incidents enabled.</Note>

### Response modes

| Mode | Behavior |
| - | - |
| **Proactive** | Devin posts a likely root cause, then keeps investigating, posts updates, and maintains an incident report |
| **One-shot RCA** | Devin posts a likely root cause with a confidence level, then stops. For incidents where you click **Start proactive mode**, it keeps investigating as an incident teammate |

### Incident settings

| Setting | What it does |
| - | - |
| **Auto-join incident channels** | Adds Devin to new channels that match the prefix. Only public channels can be auto-joined |
| **Allow non-organization users participation** | Lets people in other organizations under the same account interact with Devin in incident channels |
| **Agent mode** | The mode for the main incident session, which runs for the whole incident: it posts updates and maintains the incident report |
| **Initial investigation mode** | The mode for the first investigation, which produces the likely root cause when the incident starts |
| **Thread session mode** | The mode for sessions that answer questions and requests in channel threads |
| **Run as** | The identity incident sessions run as |
| **Connectors** | The MCP connectors incident sessions can use |
| **Runbook** | Optional. Instructions provided to every incident session |

The **Incidents** tab lists each incident with a link to its session. Use **Stop** to end Devin's involvement in an incident, and **Resume** to bring it back.

## Reports

A report gives you insights to act on from your oncall, on a recurring schedule: alert thresholds to tune, classes of issues that keep recurring to fix, and what is driving your alerts, bugs, and incidents.

Each report also summarizes the issues identified and fixed, incidents investigated, and PRs Devin opened and merged.

Use reports for a weekly on-call handoff, or scope one to your team, such as Billing or Infrastructure.

### Create a report

1. On the **Oncall** page, open the **Reports** tab and click **Create report**.
2. Enter a **Name**.
3. Under **Responders match**, choose which responders the report covers, either by selecting individual responders or by metadata.
4. Under **Incidents match**, describe which incidents belong in the report. Leave it blank to exclude incidents.
5. Set the **Schedule**, and optionally add **Instructions** describing what the report should contain.
6. Under **Notifications**, choose a Slack or Microsoft Teams channel to post the report to.
7. Click **Create report**.

Reports are shared, so they run as a service user rather than as an individual. Open a report to read past runs, or click **Generate report** to run it now.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.