The engineering memory layer for GitHub

Understand any codebase. Change it without breaking things.

Variorum learns your repository: the code, the docs, and the decisions behind them. Ask how something works, or what a change is likely to break, and the answer comes back with citations you can click. It keeps your docs in sync and flags risky changes, so the context doesn't walk out the door when people do.

Free to start · Lives in your existing GitHub workflow · Variorum proposes, you decide

variorum.app/dashboard
11
Repos
34
Docs
6
Risks
Analysis activity
mediumdocs/auth.md drifted from src/auth.py
Every answer cited Human in the loop, never auto-merges No new tool to learn Free to start

The problem

Your code is version-controlled. Your team's knowledge isn't.

The context around the code lives in scattered commits, old pull requests, and people's heads. How it fits together and why it's built this way rarely gets written down, and what does exist fades as the team grows and people move on.

New code is a maze

Every change starts with an hour of grep, git blame, and “who owns this?” in Slack.

Docs go stale

The code moved on; the README still describes last quarter's architecture.

The “why” disappears

The reason behind that workaround left with the engineer who wrote it.

Risk is invisible

You only find out a change was risky, untested and owned by one person, after it has already broken.

Ask the codebase

Get an answer, not a search result

  • Ask in plain English. Variorum answers from the actual code, docs, PRs, and decisions.
  • Every answer cites its sources; click a citation to jump to the exact lines on GitHub.
  • Onboarding drops from days of spelunking to minutes.
Engineering memory

> How does rate limiting work?

Requests are throttled in rate_limit_middleware, added to move abuse protection off each route after the incident in PR #182.

core/ratelimit.py:24 DecisionPR #182

Plan a change

Know what you'll break before you touch it

  • Describe the change; get the files to edit, how risky each is, and who to loop in.
  • See the decisions that explain today's design, the docs that will drift, and the tests you're missing.
  • The review-time surprise becomes a pre-work checklist.
Plan a change
Before you start: exports are high-churn and untested; loop in Sam before changing them.
src/exports.py🔴 hotspot · no tests
exports/ ownerSam · sole owner ⚠
Decision #7exports kept synchronous
tests to addsrc/exports.py

Right where you work

Insights on the pull request, not in another tab

  • Every PR gets an impact briefing (hotspots, owners, and missing tests) as one sticky comment.
  • Documentation drift and test gaps come back as reviewable PRs. Variorum proposes; you merge.
  • The comment updates in place as you push, so it's never a stale wall of duplicates.
On the pull request
Variorum · PR briefing
src/exports.py🔴 90 · no tests
src/report.py🟠 single-owner

📄 1 doc-drift · 🧪 1 test-risk flagged for this PR.

Meets you in Slack

The most reliable dashboard is the one you never have to open

Health only helps if someone looks. Variorum sends a weekly recap, plus the alerts that can't wait, to the channel your team already lives in. A health drop or a new critical hotspot reaches you the day it happens, not the week you next remember to check.

  • A weekly digest covering health, new drift and risk, and the top hotspots, on the day and time you pick.
  • Connect once with an incoming webhook; set the cadence per repository.
  • Send any repository's digest to Slack on demand, straight from the dashboard.
# engineering
V
VariorumApp9:00 AM

🟡 acme/payments · health 61/100 · last 7 days

• 2 new doc-drift · 1 new test-risk · 3 knowledge added · 2 single-owner modules

Top hotspots

src/exports.py (90)

src/billing.py (72)

Everything else

The rest of the memory layer

Documentation Intelligence

When a PR changes code, Variorum spots the docs it left behind and opens a fix PR, with evidence for every claim. Your docs stay honest and nobody has to babysit them.

Testing Intelligence

Each change gets a risk score and a list of scenarios that look untested. Variorum can open a test PR for review, so those gaps get caught before they ship.

Digests & alerts

A weekly recap of drift, risk, and knowledge lands in Slack; health drops and new critical hotspots page you the moment they appear.

Portfolio & expertise

See knowledge health across every repo, and who actually knows each area, so a bus factor of one turns up before that person hands in their notice.

Codebase understanding

A structural map of your files, functions, and classes, plus how the documentation relates to the code. It refreshes on its own every time you push.

Fits your PR flow

One analysis per pull request, posted where reviewers already work. No dashboards to babysit, no workflow to change.

How it works

Connected in minutes, useful the same day

01

Connect a repository

Install the GitHub App and pick the repositories you want Variorum to learn.

02

It learns your repo

Variorum maps the code, reads the docs, and ingests the history behind them. From then on it stays fresh on every push.

03

Ask and plan

Ask how anything works, or plan a change and see what it touches. It's all cited.

04

Review on your PRs

Briefings, drift fixes, and test gaps arrive on the pull request. You review and merge.

Not another coding agent

A coding assistant writes the next line. Variorum remembers the last thousand.

Variorum doesn't generate your code. It builds a durable, cited memory around it (the how, the why, and the risk) and puts that in front of you at the moment you need it. It proposes; your team always decides.

Give your team a memory that outlasts any one engineer

Connect a repository and, within minutes, start asking how it works, planning changes safely, and keeping docs honest.