Nexus multi-persona review

A single reviewer has single blind spots. Nexus runs code review as a crew: multiple AI specialist personas examine the same code in parallel, each with its own role, expertise, and opinions, and the results are aggregated into one review with per-persona comments.

How a review runs

1. You submit code (or a project artifact) for review and choose the personas — by default Maya (QA Architect), Kit (Pragmatic Developer), and Vex (Security Expert).
2. Each persona reviews independently and in parallel with its own system prompt and, optionally, its own slice of context.
3. Comments are aggregated with per-persona attribution, severity, and category; a failed persona degrades gracefully to an informational note instead of failing the review.
4. The result is a structured review you can revisit, export to PDF, and share via a link.

Personas consult, not just comment

Personas know the crew roster and defer to each other by design — Maya defers to Vex on security and to Beacon on accessibility matters. In chat mode you can talk to the crew, stream responses, and review the full history of a conversation.

Models

Nexus supports multiple AI providers and routes per review; the platform default is provider-auto with an optional model override. A provider failure on one persona does not take down the run.

Budgets

Nexus usage is metered against your plan’s monthly token budget and, for full reviews, a monthly review count. Exceeding a budget returns a clear 429 with the used/limit numbers rather than a silent failure. Trials are available so you can evaluate the crew before committing.

Where it fits

The pipeline gates (Alpha onward) run automatically on PRs. Nexus is the on-demand, deeper pass: use it when a change is risky, when you want a second opinion before opening a PR, or when reviewing test strategy rather than a single diff. The full crew is catalogued in the persona reference.