skuggi.

AI code review · GitHub-native

An AI engineering team in your shadow.

An AI engineering team for your GitHub repo. It plans the issue, implements it, reviews every PR and fixes what it finds.

Join the waiting list and your first readiness scorecard is free at launch. Metadata-only, no card, no setup.

  • Under 2 min per review
  • GitHub-native
  • Learns your repo
  • No config file

Inside the shadow

Every pull request gets this.

A real review, from Skuggi's dashboard. PR #178 on a test repository: one warning, under one credit, 1m 54s.

GitHub-native

No new tool. It works in your pull requests.

Skuggi installs as a GitHub App and posts its review on the PR, like a teammate would: a summary, findings ranked by severity, the failure traced through the code, a checklist. Your team stays in GitHub. Mention @skuggi when you want more.

Posted by Skuggi on a pull request in one of Anantys' own repositories.

PR Lens

It maps what your change touches.

Next to the findings, a review can carry a map of the change: what it adds, what it calls, what else it can break. Drawn from the code, posted with the review.

PR Lens on the same PR #178: new components outlined in green, the calls they add, the module they lean on.

Install it. That’s the setup.

No model to pick. No config to write. No card required for the first review. We pick the best model for each PR, so you don’t have to.

The pipeline, from the dashboard: planned, reviewed, then fixed.

Gets smarter over time.

Skuggi keeps your repo’s context from one review to the next. The pattern it flagged in March? It remembers how you fixed it, and holds the next PR to it.

Review #100 will be sharper than review #1. That’s the point.

Review #2 · four months ago
warningsrc/middleware/auth.py:88 — Bare except is too wide
Review #14 · today
warningsrc/api/session.py:61 — Same bare-except pattern you fixed in PR #98. Same fix applies.
remembers PR #98

The loop

One agent. Your whole maintenance loop.

Plan, implement, review, fix: one agent, one repo memory, the whole loop. All four ship together in the MVP. The agent proposes; you decide.

Plan

Point it at an issue. Skuggi reads the issue and the repository, then posts an implementation plan your team validates before any code is written.

@skuggi plan

Implement

Hand it an issue, get a draft PR back. Same agent, same repo memory: it already knows the codebase.

@skuggi implement

Review

Every PR reviewed in under 2 minutes, findings ranked by severity, posted where you already work.

@skuggi review

Fix

Turn a finding into a patch. Skuggi opens the fix as a PR, and you review it like any other.

@skuggi fix

Product questions

Straight answers about Skuggi, the AI engineering agent for GitHub teams.

Which platforms does Skuggi support?

Skuggi is a native GitHub app, our focus for now. It installs on a repo in a couple of minutes and works where you already work. Other platforms will be supported soon.

How is the model chosen?

We pick the best model for each task via OpenRouter. You don't see it, you don't pick.

Can I try it without a card?

Yes. The first week is free, and your first diagnostic is free without a card.

What is Skuggi?

Skuggi is an AI engineering agent for your GitHub repositories. It learns your codebase and helps maintain it: it plans from an issue, implements a draft PR, reviews every PR and fixes what it finds. The agent proposes; the human decides.

Be first in line when Skuggi opens.

Join the waiting list