For engineering leads

Your team ships AI-written code every day. Who's checking it?

Knitwise by Blore.AI shows how your team really works with Claude Code and Codex, then helps you agree one standard and keep it.

Get your free assessment See a sample report

Free · Read-only · Your code never leaves your CI · Open-source assessment

Knitwise — AI-Native Engineering Maturity Report

Preview

Agent setup 5/5
PR hygiene 5/5
The problem

AI adoption happened. The standard didn't.

Everyone uses AI differently

Each developer has their own setup, rules and habits.

Reviews can't keep up

AI writes more code, faster, than teams can properly review.

Leaders can't see it

You know AI is in every PR. You don't know how it's checked.

From our own codebase

We run Knitwise on the code that builds Knitwise.

PR hygiene went from 1/5 to 5/5 in three days.

How it works

See it. Fix it. Keep it.

01 · See it

A free, read-only check of how your team really ships code

Knitwise runs against your repo's own history — commits, PRs, reviews, CI — and scores six real areas of practice.

Every score links back to the evidence behind it, and the report ends with the fixes that matter most, ranked by impact.

Knitwise — AI-Native Engineering Maturity Report
Agent setup 5/5
Tests 5/5
PR hygiene 5/5
AI adoption 5/5
Review depth Not enough data (one maintainer)

Top fixes

  1. Route changes to code owners
  2. Review new dependencies in CI

Real report from the Knitwise codebase

02 · Fix it

One agreed standard, set up for you

We open a single setup pull request with the standard already built in, then align your team on it in a half-day session.

Nothing is imposed silently — your team reviews and merges the PR itself.

Open Add Knitwise standard #128 into main

In this PR

CLAUDE.md
AGENTS.md
.github/pull_request_template.md
.github/workflows/ci.yml

Applied by your admin

Branch protection on main
Required reviewers
03 · Keep it Coming soon

A weekly note when the standard starts to slip

Once the standard is in place, Keep it watches for drift and tells your team before it becomes a habit.

Weekly digest Coming soon
Agent setup
Safety checks
PR hygiene
Where it sits

Runs where your code already lives.

Your repo Where it starts
Knitwise Assess Read-only, in your CI
Report Evidence + top fixes
Setup PR Plus a half-day session
Weekly digest Coming soon
Why teams trust it

Built for teams that care about their code

Your code never leaves your CI.

Open source: your security team can read every line. View on GitHub

Works with Claude Code, Codex, or both.

Fit check

Is it for you?

Great fit

  • 5–15 developers
  • Already using Claude Code or Codex
  • Want a standard, not a mandate

Not for you yet

  • No AI coding agents in use
  • Looking for AI courses? Browse our courses
  • Want to monitor individual developers
FAQ

Questions engineering leads ask

No. It runs as a read-only GitHub Action (knitwise-dev/assess) that only makes GET calls to api.github.com from inside your own CI — your source code is never transmitted anywhere.

Six areas, each 0–5 when there's enough data to score it. Full detail in how scores are calculated.
Agent configuration
Test discipline
Review depth
PR hygiene
Safety gates
Adoption

Yes, the assessment itself runs on any plan. GitHub Free doesn't let private repos enforce branch rules, though, so required reviews and required checks aren't visible to the API there — upgrading to GitHub Pro or Team unlocks full scoring on those two areas.

Claude Code and Codex, detected from each repo's own configuration. A team using only other agents (for example Cursor or Copilot) will score low on Agent setup regardless of how well-configured they actually are.

No. Every report aggregates to team and repository level — nobody is named, ranked or scored individually.

A tailored pull request to your repo (consistent CLAUDE.md/AGENTS.md, agent permissions, a test hook, a PR template, CI checks) plus a half-day session to agree the standard as a team.

The assessment is free. Setup is quoted per team based on scope — ask us for a number, we won't guess one here.

Direct-to-main commits are counted but not fully assessed as PRs, AI use with no trace (no co-author trailer, branch name or label) is invisible, and it isn't a code-quality or security scanner. Full detail in the limitations doc on GitHub.
Early access

Be one of our first 10 teams

Free assessment plus a 30-minute walkthrough of your report, at no cost.

Get your free assessment Also an Authorized Microsoft AI Cloud Partner.
Agent setup 5/5
PR hygiene 5/5
Get your free assessment