How Code Review works
Unblocked reviews pull requests in GitHub, GitLab, Bitbucket, and Azure DevOps, including supported self-hosted editions. It can generate pull request descriptions that explain the changes and their context, post inline review comments with suggested changes, publish a commit status, approve low-risk changes, and re-review pull requests when you push new code. You can configure Code Review to fit your team’s workflow: choose which repositories and authors receive reviews, filter pull requests by branch, label, or title, exclude files, and decide whether to review drafts or follow repository instructions. See Code Review configuration for all available settings.PR Chat
Use PR Chat to talk to Unblocked directly in a pull request by mentioning@unblocked in a comment. Ask questions about the code or the reasoning behind a change, request a summary or diagram, or ask for another review focused on a specific concern. You can follow up with questions and additional context in the same conversation. Learn more about PR Chat.
Fixing review findings
Use the fix buttons on review comments to ask Unblocked to address an individual finding or the review’s findings together. You can also ask in natural language, such as “@unblocked fix this,” and include any constraints or guidance. Unblocked handles the changes through a coding task and links to the task so you can follow its progress.Risk-gated auto-approval
Use risk assessment to enable auto-approval for changes your team considers low risk. You define risk policies in the repository, and Unblocked evaluates each pull request against them. With auto-approval enabled, pull requests qualify when their risk score islow or lowest and the review finds no issues. Learn more about risk assessment.
Unblocked’s approval counts toward your branch protection requirements. You can also make the Unblocked Review commit status a required check to ensure reviews finish before merging. The status reports review completion; risk-gated approval determines whether the change qualifies for automatic approval.
Code Review Memory
Code Review Memory learns your team’s conventions and decisions from replies to Unblocked review comments. Explain how you want future reviews to handle a pattern, and Unblocked checks that guidance against the code before saving it as a rule scoped to the relevant repository, folder, module, or file type. No setup is required. Replies such as “fixed, thanks” and one-off exceptions are not saved as rules. Memory can reduce repeated comments about accepted patterns or require closer checks where your team needs them.
Code Review Memory: less noise, same recall
How Unblocked learns your team's review decisions, checks them against the code, and applies them to future pull requests.