VZU Portfolio · code-review
A separate reviewer agent reads every diff. A humans-in-the-loop override handles the corners.
A code review surface on the Netwit Agentic OS runtime. A separate reviewer agent reads every diff, with a humans-in-the-loop override for the corner cases.
Code Review is a code-review surface on the Netwit Agentic OS runtime. A separate reviewer agent — different prompt, different model, different rubric — reads every diff before it lands. The reviewer is not the writer. The rubric is configurable: security, performance, accessibility, type safety, test coverage, MCP contract. The comments are inline, line-level, with the reasoning attached. The override is logged. The review lands in the same PR thread the team is already reading. Code Review is a tool inside the VZU family, operated by Virtual Zero Unbound Inc.
What ships
Lead capability
02
Security, performance, accessibility, type safety, test coverage, MCP contract — turn the ones you need on, turn the rest off. Per repo, per team.
03
The reviewer explains the why, not just "this could be better". Reasoning attached, override logged, every comment lives in the PR thread.
04
Every diff, every comment, every override, every resolution. The audit trail of the codebase lands in the repo, not in a Slack channel.
05
A humans-in-the-loop override for the cases the agent gets wrong. The override is recorded, not silent. The audit trail stays honest.
06
GitHub, GitLab, Bitbucket — the review lands in the same PR thread the team is already reading. No new tab, no new tool, no new context switch.
The deep view
The point of Code Review is the separation. The reviewer is not the writer. Different prompt, different model, different rubric. The separation is the point — when the same agent writes the code and reviews the code, the review is a rubber stamp. The reviewer reads the diff, the surrounding code, and the rubric. The reviewer is not the writer. The override is the safety net for the corners.
The rubric is configurable. Security, performance, accessibility, type safety, test coverage, MCP contract — the team turns the ones it needs on, turns the rest off. The rubric is per repo. A regulated repo can run the security + accessibility rubric on every PR; a SaaS backend can run the performance + type-safety rubric; an open-source maintainer can run a default rubric with a humans-in-the-loop override on the final accept.
The comments are inline and line-level. The reviewer explains the why, not just "this could be better". The team can accept the comment, override the comment, or ignore the comment. Every override is logged in the review log in the repo, with the author and the rationale. The override rate is tracked per author, not to shame the team — to surface the corners where the rubric needs tuning.
The review lands in the same PR thread the team is already reading. GitHub, GitLab, Bitbucket — native, with the same line-level comments the team is used to. No new tab, no new tool, no new context switch. The audit trail of the codebase lives in the codebase. The review log is in the repo, the diff is in the repo, the override is in the repo. The reviewer agent reads the diff overnight, the maintainer reads the diff in the morning.
The process
The reviewer agent reads the diff, the surrounding code, and the rubric. It writes inline comments with line-level reasoning. The reviewer is not the writer. The separation is the point.
Security, performance, accessibility, type safety, test coverage, MCP contract — turn the ones you need on, turn the rest off. The reviewer reads the rubric before the diff. The rubric is per repo.
The team reads the comments, accepts the ones that are right, overrides the ones that are wrong. Every override is logged in the review log in the repo, with the author and the rationale.
The review log is committed alongside the diff. The audit trail of the codebase stays current. The override rate is tracked per author, per rubric, per quarter.
Where the work happens
The Code Review surface is operated out of three VZU offices. The reviewer agent reads overnight, the team reads the diff in the morning, the override is logged in the repo. That is the loop.
In production
Regulated services
With a humans-in-the-loop override on the architectural decisions. The review log is the regulator's form. The override rate dropped 60% in the first quarter. The reviewer caught three contract drifts in the API layer before they shipped.
SaaS backend
A 90% comment-acceptance rate over the first quarter. The reviewer agent caught a type-safety regression in the payments module that would have shipped to staging. The override is logged, the audit trail is honest.
Open-source maintainer
With a humans-in-the-loop override on the final accept. The review log is in the repo. The maintainer reads the diff in the morning; the agent has already read it overnight. The override rate tells the maintainer where the rubric needs tuning.
Other VZU tools
On the runtime
Code Review is a configuration of the Netwit Agentic OS runtime. The reviewer agent and the writer agent are not the same prompt. The runtime enforces RBAC at the agent level, ships the audit trail by default, and deploys in cloud, hybrid, or on-prem modes.
The VZU platform, in detail →Questions
Get in touch
Email the team. Reply in one business day. The pilot is priced, the rubric is configured, the ship date is real.