Use cases / AI Code Review
Use case
One step in the pipeline you already run. Every push to a pull request is re-checked against your rule packs and your codebase knowledge, the findings land on the pull request, and they close themselves when the code changes. A human approves the merge.
One review step, inside your CI
Your CI
Re-checked against your rule packs and your codebase knowledge. Only the diff, the symbol facts that document it, and the agent's conversation are reviewed.
Findings posted where the review happens
The team reads it where it already talks
What Uzera does on every pull request
Code Review posts the finding, Guardrails decides what counts, and Trust keeps the record.
Code Review
Add one step to the pipeline you already run. Nothing is installed on the build runner and nothing else in the pipeline changes. GitHub Actions, Azure Pipelines, Bitbucket Pipelines, Azure DevOps Server.
Guardrails
Every push is checked against the rule packs you adopted and the rules you wrote in plain English, plus what Uzera knows about your codebase.
Code Review
Push the fix and the finding is gone. Nobody marks it done. Its identity survives refactors, so it does not come back under a new line number.
Trust
Every finding and every verdict is written to a hash-chained ledger, and the pull request keeps the history of what was raised and what was resolved.
How the review step runs
One step in the pipeline you already run. The rest happens on the pull request.
Add one step to the pipeline you already run. Nothing is installed on the build runner, and nothing else in the pipeline changes.
Each push is checked against your rule packs and your codebase knowledge. Only the diff, the symbol facts that document it, and the agent's conversation are reviewed. Never a clone of your repository.
Findings close themselves when the code changes. A human approves the merge, and every allow and deny is recorded on the hash-chained ledger.
Where it fits
Before the budget conversation
Uzera doesn't replace it. It makes the agent you already pay for follow your rules, and shows you what it cost, which is how you justify the spend you have rather than adding to it.
No. It runs before a reviewer opens the pull request. Findings name the rule and the line, so the reviewer spends their time on the design of the change, not on conventions.
No clone of your repository is ever taken. Three things are sent: the diff of the change under review, the names and locations of the symbols it touches, and the agent's conversation for that session. Nothing else. The index of your codebase stays on your machine.
The next push is re-checked against the same rules. A finding that no longer applies closes itself, and re-runs never re-bill.
Governance for your coding agents
AI writes your code. You stay in charge.