Demonstrating PR Agents

00:00 Pull request agents are structurally different from IDE and terminal agents, and it comes down to one word, asynchronous. You don’t watch a PR agent work in real time.

00:11 Instead, the agent often triggers automatically when a pull request is opened or updated.

00:17 This workflow operates on shared branches, the ones visible to everyone on the team. The agent flags potential bugs, style violations, and logic issues, but a human reviewer always makes the final call on whether to merge.

00:33 PR agents act as a safety net before merging, not some live tool that you use while coding.

00:40 Tools like CodeRabbit and GitHub Copilot code review support this workflow on GitHub, GitLab, and Bitbucket.

00:48 Now let’s see GitHub Copilot code review in action. Remember the code example that GitHub Copilot fixed? Let’s use the original version, the one that had the bug, to create a PR and see what GitHub Copilot Code Review says about it.

01:04 The workflow in practice looks like this. You open a pull request and add Copilot as a reviewer. Sometime later, the agent will post a review.

01:16 It might catch an unhandled edge case, flag a missing test, or suggest a clearer approach.

01:22 Copilot’s review comments behave like review comments from humans, so here you can see what GitHub Copilot came up with. It left a pull request overview, what this PR is about, and here it left a comment.

01:36 And it caught the bug that we were already aware that existed in the code. So we can see the suggested changeset.

01:45 So the suggestion that it gives looks okay, so let’s do Commit suggestion and Apply Suggestion, so it will fix the bug. In case it wasn’t correct, you can always reject it.

01:58 At this point, the PR is ready to merge. It is worth noting that on many teams, repository-level AI tools are approved or blocked at the organizational level.

02:10 Privacy decisions here often happen above the individual developer, so check your team’s policy before connecting a PR agent to a shared repository.

02:20 In the next lesson, you’ll look at the workflow type with the most autonomy of all, cloud agents.