This is the second post in our AWS DevOps Agent series. In our first blog, we have seen how to set up DevOps Agent & integrated with cloudwatch alarms
An investigation agent is only as useful as the tools it's wired into. In the last post we covered CloudWatch alarms and generic webhooks — the "how it hears about problems" layer. This post covers the "where it talks to your team and your code" layer: Slack for notifications, Jira for ticketing, and GitHub for the code side of things. All three connect differently, and knowing why upfront saves you some head-scratching.
In this blog post, we are going to integrate DevOps agent with Slack, JIRA & Github.
Slack — Built In
Slack is a native integration — no MCP server, no separate app to host. Once connected, AWS DevOps Agent posts key findings, root cause analysis, and generated mitigation plans directly into whichever channels you choose.
Prerequisite: You may need Slack Admin access to install AWS DevOps agent app.
In some cases you may not have admin rights to add this integration, in that scenario, you need to ask Slack admin to install AWS DevOps Agent App for you.
Setup is two steps:
- Register the integration — Go to respective agent space, find Slack under Communication and choose Register. You'll be redirected to Slack to authorize the app — install it per workspace.
If no workspace is added, click Register, it will take you to Slack login screen.
- Associate channels with your Agent Space — from your Agent Space's Capabilities tab, go to Communications → Slack → Add Slack, and enter the Channel ID for each channel that should receive updates.
In Slack, you can invite AWS DevOps agent or you can also add it under Agent & Apps tab
Now we have everything in place, lets have short demo.
DEMO - CloudWatch Alarm Investigation Notification
In last blog, I have investigation for CloudWatch alarm Cloudfront5xx errors, as soon as investigation started by devops agent I get notification in slack, with timelines, root cause & investigation gaps. No need to login in AWS Console or DevOps agent WebApp
So without switching screens, you get notification in one place.
Jira — Via the Atlassian Rovo MCP Server
Jira isn't a built-in integration, so it connects through Atlassian's own hosted Rovo MCP Server — the same server that also covers Confluence, so connecting Jira gets you Confluence
Step 1 : Configure Rovo MCP in DevOps agent, Agent Space -> Capabilities -> MCP Server
| Field | Value |
|---|---|
| Name | jira-mcp |
| Endpoint URL | https://mcp.atlassian.com/v1/mcp |
| Enable Dynamic Client Registration | ✅ Enabled |
Approve access in Jira if you already have admin access for Jira.
In case if you face an issue - you may need to add DevOps Agent Domain in Atlassian Admin under Rovo MCP Server -> Domain. URL : https://*.api.aws/**
Step 2 : Permission for an Agent Space
Once you have MCP access is in place, select which MCP tools you can use. Go for permission which are required only, for demo part I have given access to create tickets for investigations.
Step 3 : Verification
After completing the setup, verify that the integration is working:
- Open DevOps agent WebApp
- I will be using below prompt :
Check for stale/unused resources and create respective Jira ticket under project "cg-helpdesk".In few minutes, it will create Jira tickets
Now to go one step further, let's have Skill in place to create tickets , in DevOps agent WebApp go to Knowledge -> Skills -> Custom Skills. You can create custom skill with below description to have ticket in JIRA, obviously modify it as per your need.
Please note that, its advisable to add Guardrails to avoid data-leakage, ticket noise. As progress happens, you can always modify this skill.
---
name: cloudwatch-alarm-jira-ticket
description: Automatically create Jira tickets in the _my-jira-app_ project when investigating CloudWatch alarms. Use this skill during CloudWatch alarm triage or investigation to log findings, affected resources, severity, and recommended actions as a structured Jira task. Targets the Incident-Triage and Incident-RCA agent types.
---
# CloudWatch Alarm Investigation — Jira Ticket Creator
## When to Use
- A CloudWatch alarm has fired and an investigation is underway
- An investigation has produced findings related to a CloudWatch alarm
- The user asks to "create a Jira ticket", "log this alarm", or "track this in Jira" during an alarm investigation
- Post-investigation wrap-up where findings need to be tracked for remediation
## Prerequisites
- Jira Cloud ID: `1231412535345435345` (demo.atlassian.net)
- Jira Project Key: `HELP` (my-jira-app)
- Issue Type: `Task`
- Use the `jira-mcp_createJiraIssue` tool (available via the jira-mcp MCP server)
## Severity → Priority Mapping
| CloudWatch Alarm Severity | Jira Priority |
|---|---|
| CRITICAL | High |
| HIGH | High |
| MEDIUM | Medium |
| LOW | Low |
| UNKNOWN | Medium |
## 🚧 Guardrails
These rules are mandatory and must be checked before any ticket is created.
### 1. Severity Threshold — Avoid Ticket Noise
**Do not create a Jira ticket for investigations whose alarm severity is below your team's configured threshold.**
- Default threshold: **MEDIUM and above** (i.e. skip LOW severity alarms).
- If the alarm severity is LOW, log a note in the investigation findings stating: "Ticket creation skipped — alarm severity is below the configured threshold (LOW). Adjust the threshold in the skill if this should be tracked."
- Teams should adjust this threshold to match their operational standards.
### 2. One Ticket Per Investigation — No Duplicates
**Never create more than one Jira ticket per investigation.**
- Before creating a ticket, check whether one already exists for this investigation by searching for the investigation ID in the ticket labels or description.
- If a ticket already exists, add a comment to the existing ticket instead of creating a new one.
- If you cannot verify (e.g. search is unavailable), note this in the investigation findings and skip creation rather than risk a duplicate.
### 3. Jira Failure — Surface, Don't Silently Drop
**If the Jira create-issue call fails or Jira is unreachable, do not silently discard the failure.**
- Record the failure explicitly in the investigation notes/findings with:
- The error message or HTTP status returned
- The full ticket content that was attempted (summary + description)
- Instruction: "Jira ticket creation failed — a human must file this ticket manually in cg-helpdesk (HELP)."
- Do not retry more than once automatically.
### 4. No Personal Information in Tickets
**Never include personal details in any Jira ticket field (summary, description, labels, comments).**
- Do not include: personally identifiable information (PII).
- If an alarm or log message contains PII (e.g. a username in a CloudWatch dimension), redact it — use a placeholder like `[REDACTED]` or refer to the resource by its ARN/ID only.
---
GitHub — Native Pipeline Integration
GitHub is a first-class, native integration, and it does more than notify — it lets AWS DevOps Agent correlate deployments with incidents, and (in preview) run release readiness code reviews and automated testing directly against pull requests.
To have integration in place :
Step 1 : Go to respective agent space -> Pipeline -> Add Source -> Register Github
Step 2 : Enter details as per your org or user , you can also directly hit Submit , it will take you to your Github org permission page,
Select Specific repo only :
Step 3 : Configure code review & testing. Here I'm configuring only code review
When Code Review enabled for a repository, DevOps Agent automatically runs a Release readiness code reviews each time a pull request is opened or updated. Review findings appear as inline comments on the pull request. This is enabled by default for all connected repositories.
DEMO
I will change my webapp to orange theme and will create pull request, this triggers automatic code-review in devops agent
From the above screen shot we can see, devops agent able to identify which code change are done and also recommendation part.
Key Takeaways
- Slack posts findings and mitigation plans straight into your channels — one workspace install, then channels added per Agent Space.
- Jira connects through Atlassian's Rovo MCP Server, add customer skill as per your need to have ticket creation.
- GitHub is a native integration that goes beyond notifications — it correlates deployments with incidents and, in preview, runs automated code review and testing on every pull request
I hope this blog gave you a solid idea and a starting point for integrating with all three providers.
Next up in this series: cross-account monitoring, so a single Agent Space can see your whole AWS footprint.
Top comments (0)