JANUS — YOUR AI BACKLOG ENGINEER FOR AZURE DEVOPS

Most AI tools make you leave Azure DevOps to get things done. Janus works the other way around — it lives inside ADO, connects to 8 different AI providers, and handles everything from polishing a single work item to converting a full Business Requirements Document into a production-ready backlog. No tab switching, no copy-pasting, no vendor lock-in.

Janus supports all native ADO work item types across all four process templates (Scrum, Agile, CMMI, Basic) — plus custom work item types.

Your API keys are stored in ADO user storage and never transmitted without PII sanitization.

Reach us anytime through the Support panel built directly into Janus — accessible via the floating button in the bottom-right corner of every work item. It's enabled by default and can be hidden from Settings → General if you prefer a cleaner interface.

QUICK START

1. Install Janus from the Visual Studio Marketplace into your Azure DevOps organization

2. Go to Project Settings → scroll down to the Extensions section → click 📝 Janus AI Assistant ✨

   

3. In the General tab, activate an AI provider — enter your API key, select a model, and click Save

4. Open any work item — Janus actions appear in the work item form panel and the right-click context menu

That's it. No agents to deploy, no external dashboards to configure.

Note: The only required setup is adding an API key in the General tab. All standard ADO work item types are enabled by default, and project administrators have access out of the box — no need to configure the Work Items or Permissions tabs to get started. See the Setup section below for advanced configuration options.

FOUR CAPABILITIES. ONE EXTENSION.

① IMPROVE WORK ITEMS

Janus reads your existing work item and rewrites it — title, description, and acceptance criteria — using the AI model of your choice. Before anything is saved, you review every change through a word-level visual diff. You decide what stays.

• Rewrites title, description, and acceptance criteria in a single pass
• Word-by-word diff so you see exactly what changed and why
• Attach screenshots, mockups, or documents — vision-capable models analyze them directly
• Add your own context and regenerate as many times as you need
• Works with all native ADO work item types and custom types

② DECOMPOSE WORK ITEMS INTO CHILDREN

Give Janus a high-level Epic or Feature and it produces a fully typed, properly linked child hierarchy — automatically adapted to your project's process. It checks what already exists so it never generates duplicates.

• Respects Scrum, Agile, CMMI, and Basic process templates
• Infers the correct child type at each level of the hierarchy (Epic → Feature → User Story / PBI → Task)
• Skips children whose titles already exist under the parent
• Assign team members and create all items in one click
• All children are auto-linked to the parent before saving

③ GENERATE TEST CASES — IN YOUR FORMAT

Janus generates complete, step-by-step test cases from a work item's title and description. Pick the format that matches how your team tests — and save straight into any Test Plan and Test Suite in ADO.

Three formats built in:

• Pass a screenshot or UI mockup — vision models generate tests from images
• Edit step actions and expected results inline before saving
• Reorder, add, or delete individual steps
• Save to an existing plan/suite or create new ones on the spot
• Refine and regenerate with additional instructions

④ BRD → BACKLOG (Beta)

This is Janus's most distinctive capability. Drop in a PDF, DOCX, or TXT business requirements document and Janus runs a six-stage pipeline that classifies sections, extracts requirements, and assembles a complete, linked work item hierarchy — ready to land directly in Azure DevOps.

What gets generated: Epics · Features · User Stories · Tasks · Test Cases · Business Rules

The pipeline:

Reading → Classifying → Generating → Linking → Validating → Done ✓

• Accepts PDF, DOCX, and TXT up to 30 MB
• Drag-and-drop upload with real-time stage tracker
• PII is sanitized before the document reaches any AI model — and again on the way back
• Triggered directly from the ADO work item context menu

ADD SUGGESTIONS — GUIDE THE AI BEFORE IT ACTS

Before running any generation — improving a work item, decomposing into children, or generating test cases — you can open the Add Suggestions panel to give the AI additional context. This is optional, but powerful.

• Add free-text instructions, constraints, or style preferences
• The available fields adapt to the work item type — an Epic gets different options than a Task or a Bug
• If your selected model supports vision (GPT-4o, Claude 3.5+, Grok Vision), you can attach an image or file directly to the context — mockups, wireframes, error screenshots, documents
• Your suggestions are sent along with the work item content to the AI — giving it the extra signal it needs to generate exactly what you want

This panel is the difference between "AI does its best guess" and "AI does what you actually need."

BRING YOUR OWN AI — 8 PROVIDERS SUPPORTED

Janus does not tie you to a single AI vendor. Connect the provider your organization already uses, or switch models at any time without reinstalling anything.

Vision-capable models (GPT-4o, Claude 3.5+, Grok Vision) can process attached images, PDFs, and documents as part of any generation request.

SETUP IN THREE TABS

Open the Janus Settings Hub from your Azure DevOps project settings.

General — Connect your AI provider

Enter your API key, pick a provider, and select a model. Azure OpenAI users add the deployment name and endpoint URL. This is where you activate Janus — without a configured provider, no generation will run.

The General tab also includes Extension Preferences where you can configure your preferred AI response language and toggle the floating support button that appears in the bottom-right corner of work items (enabled by default).

Work Item Types — Control what Janus touches

Enable Janus per work item type. Configure field mappings and default values for custom templates so Janus always writes to the right fields. Supported for all native types (Epic, Feature, User Story, PBI, Task, Bug, Issue, Test Case, Requirement) and any custom type in your organization.

Permissions — Decide who can use Janus

Grant or revoke access per team member. Use the toggle-all switch for fast rollout or lockdown. The member list stays in sync with your ADO team automatically.

WHAT SETS JANUS APART

SUPPORTED WORK ITEM TYPES

All native ADO work item types are supported:

Epic · Feature · User Story · Product Backlog Item (PBI) · Task · Bug · Issue · Requirement · Test Case · Change Request · Risk · Review · Impediment · any custom type

All four ADO process templates supported:

DARK MODE SUPPORT

Janus works seamlessly in both Azure DevOps light and dark themes.

PRIVATE NETWORK / CORS ERRORS (AZURE OPENAI WITH PRIVATE ENDPOINTS)

If your organization uses Azure OpenAI with Private Endpoints, users may encounter browser errors such as:

"Permission was denied for this request to access the 'local' address space."

Or a "Forbidden" error appearing in the Janus panel.

Root Cause: Starting with Chrome and Edge 131+ (December 2024), browsers enforce Private Network Access (PNA) security. Extensions hosted on public origins (like the Azure DevOps Marketplace) are blocked from making requests directly to private IP addresses, even if those requests are authorized by the user. This is a browser-level security change — not a Janus bug.

Three solutions — no changes to Janus required:

Option 1 — Enterprise Browser Policy (fastest to deploy)

Deploy the InsecurePrivateNetworkRequestsAllowedForUrls Chrome/Edge policy with your Azure OpenAI endpoint URL. This can be pushed via Group Policy (GPO) or Microsoft Intune.

{
  "InsecurePrivateNetworkRequestsAllowedForUrls": [
    "https://<your-resource>.openai.azure.com"
  ]
}

Option 2 — Azure API Management Proxy (best long-term)

Place an Azure API Management instance with a public endpoint in front of your private Azure OpenAI resource. Configure Janus to point to the APIM gateway URL instead of the private endpoint. APIM handles auth forwarding and routing internally, and the browser only sees the public APIM URL.

Option 3 — Firewall Allow List

In the Azure Portal, navigate to your Azure OpenAI resource → Networking → Selected Networks and Private Endpoints. Add your organization's public IP address ranges to the allow list. This lets browsers connect to the resource without hitting the PNA restriction.

Which option to choose? If you need a fast fix for a small team, Option 1 is the quickest. For organizations with strict network policies, Option 2 (APIM proxy) is the most scalable. Option 3 works if your users have predictable public IPs.

A NOTE ON AI-GENERATED CONTENT

Janus uses AI to generate suggestions — titles, descriptions, acceptance criteria, test cases, children, and backlogs. AI results should always be reviewed by a human before being accepted. The generated content is a starting point, not a final decision. Janus makes it easy to review, edit, and reject anything before it touches your work items.

ONE LAST THING

Janus was built because backlog grooming was eating too much of the sprint. Every feature in it exists to give that time back to the people actually building things.

If something doesn't work the way you expect — or works better than you hoped — tell us.
The Support panel is inside Janus, one click away.