Repository health analysis

Take a reading
of your
codebasecodebase.

RepoQ reads your GitHub repositories, scores seven dimensions of health, and lists what it finds — worst first. One number you can act on, and the evidence underneath it.

Connect a repository See what it measures

Sample readout

acme/payments-api

main · 4f1c9ab

Completed

62Needs work

Code quality71
Test coverage38
Documentation54
Security45
Maintainability68
Complexity59
Duplication82

CRITICAL 2HIGH 5MEDIUM 9LOW 4

Two hardcoded credentials and an unbounded query

Input: Any GitHub repository you can read
Scored: 7 dimensions, 0–100
Findings: Severity + priority + file and line
Review: AI summary and ranked recommendations
History: Every run kept and comparable
Hosting: Self-host it; your code stays yours

Seven dimensions, every run

The overall score is a weighted composite of these seven. Each one is reported on its own, so a repository that is well tested but undocumented reads differently from one that is neither.

Code quality

Naming, structure, error handling and the patterns the codebase actually follows.

Test coverage

How much of the source has tests behind it, and where the gaps sit.

Documentation

READMEs, inline comments and whether a new engineer could find their way in.

Security

Injection paths, exposed secrets, unsafe dependencies and weak authentication.

Maintainability

Coupling, module boundaries and how expensive the next change will be.

Complexity

Function length, nesting depth and the branches nobody wants to touch.

Duplication

Repeated logic that will drift apart the moment one copy is fixed.

∑

Overall health

The single 0–100 figure on the dial, weighted so security and quality move it hardest. Tracked per run, so you can see whether it’s going the right way.

Three artefacts per analysis

A ranked list of findings

Each with a severity, a priority, and the file and line it lives on — linked to the exact commit that was analysed.

CRITICAL

Database URL committed in config

src/db/client.ts:14

HIGH

Unbounded query on the orders table

src/orders/repo.ts:88

MEDIUM

Errors swallowed without logging

src/lib/retry.ts:41

An AI code review

A written summary of the codebase’s condition and ranked recommendations, with the model that produced it named on the record.

“The service layer is well separated, but error handling is inconsistent and the reconciliation path has no tests. Address the credential exposure first.”

✦ AI generated6 providers supported

A history you can compare

Every run is kept with its branch, commit and duration, so the score becomes a trend instead of a snapshot.

main · 4f1c9abAug 2462

main · 21b8fe0Aug 1757

main · 9cd4a13Aug 0951

Three steps to the first reading

Sign in with GitHub

One OAuth handshake. RepoQ reads only the repositories you choose to connect.

Connect a repository

Pick it from your GitHub account or paste owner/repo. Metadata is pulled in for you.

Run the analysis

A minute or two later the dial, the findings and the review are on the repository page.

Take a readingof yourcodebasecodebasetest coveragesecuritymaintainabilitycomplexitydocumentationcode quality.