Overview
Scrubby is a codebase intelligence layer. It learns the architecture, conventions, and file relationships of a connected repository, then makes that knowledge available to AI editors and to every pull request.
What Scrubby learns about a repository
| Layer | What it is |
|---|---|
| Domains | The architectural areas of the codebase (for example, "Authentication" or "Billing") and which files belong to each one. |
| Domain connections | Which domains are coupled to which others, and how strongly, so a change in one area surfaces the other areas it's likely to affect. |
| Conventions | The naming, structure, testing, and error-handling patterns the codebase actually follows, extracted from real code rather than documentation. |
These layers are built automatically the first time a repository is indexed and are kept current as new commits land. See Your First Index and Indexing.
Where the knowledge is used
Inside an AI editor
Any code editor can connect to Scrubby's MCP server. Once connected, the editor can look up a file's domain and conventions, check a set of changed files for gaps, and report back what it found. See Set-Up and the integration page for your editor.
On every pull request
The Scrubby GitHub App reviews each pull request against the same domains, conventions, and connections, and posts its findings as PR comments and inline suggestions. See PR Reviews.
Getting set up
- Sign in to create an account.
- Connect an editor and a repository.
- Run your first index.
A repository must have an active trial or paid subscription to be indexed.