Azure DevOps PR Review

Review Azure DevOps pull requests directly inside Visual Studio Code. Open a pull request from its URL, inspect changed files and review threads next to the code, and draft comments with VS Code's Language Model API, typically powered by your GitHub Copilot Chat subscription.

Azure DevOps gives you the standard pull request building blocks such as comments, threads, and approvals, but it still does not give you a strong built-in AI pull request review workflow inside VS Code. This extension fills that gap: open a pull request, click Run AI Review, go through the suggested review findings and comments, edit what you want to keep, and then publish the final review back to Azure DevOps without leaving the editor.

Why This Extension

Azure DevOps pull request review often pulls you out of the editor and into the browser. This extension keeps the review workflow inside VS Code so you can:

• open a pull request without leaving your workspace
• inspect diffs and discussion threads close to the source
• reply to review threads from inside the editor
• generate draft review comments with AI, then edit them before publishing
• use your existing Github Copilot subscription

Features

• Sign in with VS Code's built-in Microsoft authentication provider
• Connect to an Azure DevOps organization without a personal access token
• Open an Azure DevOps pull request from its URL
• List active pull requests for the current repository
• Browse pull request overview, changed files, and review threads in a dedicated ADO PR view container
• Open changed files in a diff editor backed by pull request revisions
• Reply to threads, resolve them, or reopen them
• Post summary comments
• Load repo-specific review guidance from .ado-pr-review.json
• Run AI review for the whole pull request, the current file, or the current selection
• Choose which Copilot model to use for AI review
• Edit AI-generated drafts in a dedicated draft panel before publishing

Requirements

• Visual Studio Code 1.98.0 or newer
• Access to an Azure DevOps organization
• A local workspace that matches the Azure DevOps repository for repository-aware features such as pull request listing and file navigation
• GitHub Copilot Chat installed and enabled if you want AI review features

Quick Start

1. Install the extension in VS Code.
2. Run Azure DevOps PR: Set Organization URL.
3. Enter your organization URL, for example https://dev.azure.com/contoso.
4. Run Azure DevOps PR: Connect.
5. Sign in with your Microsoft account through VS Code.
6. Run Azure DevOps PR: Open Pull Request.
7. Paste a pull request URL such as https://dev.azure.com/<org>/<project>/_git/<repo>/pullrequest/<id>.

After the pull request loads, the ADO PR activity bar view shows:

• Overview for metadata and AI readiness
• Pull Requests for active pull requests in the current repository
• Files for changed files
• Threads for review discussions

Authentication

This extension uses VS Code's built-in Microsoft authentication flow.

You do not need:

• a personal access token
• a custom Entra app registration for each user

After sign-in, the extension validates that the selected Microsoft account can access the configured Azure DevOps organization.

The extension expects an organization URL such as:

https://dev.azure.com/contoso
https://contoso.visualstudio.com

If you go straight to Open Pull Request, the extension can infer the organization from the pull request URL and store it automatically.

AI Review

AI review is optional.

When enabled, the extension uses VS Code's Language Model API to send pull request context to the model you select. In most setups, this means GitHub Copilot Chat.

You can run AI review at three levels:

• entire pull request
• current file
• current selection

The generated comments open as editable drafts before anything is published to Azure DevOps.

AI Review Prerequisites

• GitHub Copilot Chat is installed in the same VS Code environment
• At least one language model is available in VS Code
• The selected model has been approved for use if VS Code requires consent

Model Selection

• Run Azure DevOps PR: Select AI Model to choose the Copilot model used for AI review
• The extension stores the exact LanguageModelChat.id in adoPr.review.preferredModelId
• Choose Automatic to clear that setting and go back to automatic model selection
• If the configured model is unavailable in the current VS Code window, AI review is blocked until you select another model or clear the setting

Example:

{
  "adoPr.review.preferredModelId": "copilot:gpt-5.4"
}

The value above is only an example. Use Select AI Model to see the exact model ids available in your VS Code session.

Review Guidance File

The extension can load repository-specific review instructions from .ado-pr-review.json in the workspace root.

Example:

{
  "version": 1,
  "reviewInstructions": [
    "Prioritize correctness, security, and data handling issues.",
    "Avoid low-value style comments unless they hide a real risk."
  ]
}

If the file is missing, Azure DevOps PR: Open Review Config can create a starter config for you.

Commands

• Azure DevOps PR: Connect
• Azure DevOps PR: Disconnect
• Azure DevOps PR: Set Organization URL
• Azure DevOps PR: Open Pull Request
• Azure DevOps PR: Open Selected Pull Request
• Azure DevOps PR: Refresh Pull Request
• Azure DevOps PR: Refresh Pull Request List
• Azure DevOps PR: Run AI Review
• Azure DevOps PR: Select AI Model
• Azure DevOps PR: Open Review Config
• Azure DevOps PR: Show Draft Suggestions
• Azure DevOps PR: Publish Draft Comment
• Azure DevOps PR: Reply To Thread
• Azure DevOps PR: Resolve Thread
• Azure DevOps PR: Reopen Thread
• Azure DevOps PR: Post Summary Comment
• Azure DevOps PR: Open Changed File
• Azure DevOps PR: Open Thread Location

Settings

Example:

{
  "adoPr.azureDevOps.organizationUrl": "https://dev.azure.com/contoso",
  "adoPr.review.maxFilesForAi": 20,
  "adoPr.review.maxDiffChars": 18000,
  "adoPr.review.preferredModelId": ""
}

Typical Workflow

1. Connect to your Azure DevOps organization.
2. Open a pull request from its URL.
3. Inspect the overview, changed files, and existing review threads.
4. Open a changed file in the diff editor.
5. Reply to a thread or post a summary comment if needed.
6. Run AI review on the pull request, a file, or a selected code region.
7. Edit the generated drafts.
8. Publish only the comments you want to keep.

Privacy and Data Flow

The extension interacts with two kinds of services:

• Azure DevOps REST APIs to load pull request metadata, diffs, files, and threads, and to publish replies or comments
• VS Code language model providers, such as GitHub Copilot Chat, when you explicitly run AI review

If you use AI review, pull request content such as file paths, diffs, and review instructions may be sent to the selected model provider through VS Code's Language Model API.

If you do not run AI review, the extension does not send pull request code to a language model.

Local Development

Build and Verify

npm install
npm run build
npm test
npm run test:extension
npm run package:vsix

This produces:

dist/vscode-ado-pr-extension.vsix

Install the VSIX

1. Open VS Code.
2. Open the Extensions view.
3. Click ....
4. Choose Install from VSIX....
5. Select dist/vscode-ado-pr-extension.vsix.

Manual Smoke Test

Open a local repository in VS Code that corresponds to the Azure DevOps repository you want to review.

Then run:

1. Azure DevOps PR: Set Organization URL
2. Azure DevOps PR: Connect
3. Azure DevOps PR: Open Pull Request
4. Paste a real pull request URL, for example https://dev.azure.com/<org>/<project>/_git/<repo>/pullrequest/<id>

Expected result:

• ADO PR appears in the Activity Bar
• Overview shows organization, connected account, title, status, files, and threads
• Pull Requests lists active pull requests for the current repository and lets you open one directly
• Overview shows AI readiness, including whether a model is available, whether consent is still needed, or whether AI is unavailable
• Files lists changed files
• Threads lists current review threads
• opening a changed file opens a diff editor

Tests

Run unit and integration tests:

npm test

This covers:

• pull request URL parsing
• organization URL normalization
• repo review config validation
• AI draft parsing
• Azure DevOps client request and response mapping
• thread mutation payloads
• organization access validation

Run extension host tests:

npm run test:extension

This launches a real VS Code test instance and verifies activation plus core command registration. It opens a GUI VS Code process, so it needs a desktop-capable environment.

Packaging

npm run package:vsix

Troubleshooting

Connect fails before sign-in

Check that your organization URL looks like one of these:

https://dev.azure.com/<org>
https://<org>.visualstudio.com

Connect succeeds but Azure DevOps calls return 401 or 403

Usually this means one of:

• the Microsoft account does not have access to the Azure DevOps organization
• the organization URL is wrong
• the requested scopes are not accepted in your environment
• the pull request belongs to a different Azure DevOps organization than the one currently configured

No AI model is available

Make sure GitHub Copilot Chat is installed and enabled in the same VS Code environment. The Overview view reports whether AI is ready, waiting for consent, or unavailable.

Comments do not land on the exact line

The extension uses conservative fallback behavior when exact Azure DevOps diff anchoring is unclear. File-level and summary comments are usually the easiest way to confirm end-to-end behavior in a new environment.

Known Limitations

• One Azure DevOps organization is active per workspace
• One pull request is active at a time
• Large pull requests are truncated in AI review flows to keep request size bounded
• Exact Azure DevOps line anchoring is not guaranteed for all diff shapes

Feedback

If you are testing early versions of the extension, feedback is most useful when it includes:

• the pull request URL pattern you used
• whether the workspace repository matched the Azure DevOps repository
• whether the issue was in authentication, file navigation, threads, or AI review
• whether the failure was reproducible