Building meters for the engineering risks nobody can see
Kodebeat turns twenty years of building systems into open-source instruments, each with a paper explaining its numbers.
Areas
Every area starts the same way: install open-source instruments, get a baseline, and only then change anything.
Secure development, from prompt to production
Developers use AI agents and builds pull a thousand dependencies. The rails go at both ends: a gateway that meters and redacts AI traffic, jails and per-command risk scoring for the agents, and a hardened supply chain with signed images, SBOMs and no secrets left in git history.
Cloud efficiency and GreenOps
Cost, energy and carbon per team and per workload, measured directly on the cluster. The emissions figures come with a documented methodology.
Resilience proved rather than assumed
The backup restored on a schedule, the rewrite proved on mirrored traffic before it serves any, the evidence captured before the fix destroys it, and requests rightsized from what the cluster actually used. Each claim arrives as a dated run somebody else can check.
What the codebase is worth before you bet on it
A model wrote most of the diff, and instincts trained on human mistakes do not fire on it. Repository scoring with every factor shown and nothing counted that could not be checked, architecture documentation computed from the tree rather than written, and review procedures for code no human wrote.
The toolbox
All public, falling into groups. Each one exists because a real question kept going unanswered.
AI infrastructure
What an agent can reach, what it costs and what it just tried to run: containment, gateways and command risk scoring.
How agents stay contained →GreenOps
Energy and carbon measured on the cluster, carbon-aware placement, carbon budgets in CI, and a paper behind each part.
The measurement stack and its papers →Security & supply chain
SBOM diffs, hardened Dockerfiles, unsigned images blocked, RBAC drift and an inventory of everything that expires.
What went into the build →Cloud cost
What moved the bill, the resources nobody owns, least privilege from real usage, a reversible spend killswitch and CI toil priced in euros.
Where the money goes →Platform & SRE
Incident evidence collected, restores verified on a schedule, rightsizing as YAML, architecture drift and rewrites proved on mirrored traffic.
Tools for the 2am questions →Repo intelligence
Repository scores with every factor shown, review sheets for code no human wrote, and a hostile second opinion from a coding assistant.
What your codebase tells you →Teaching & writing
A free GreenOps book, a developer learning path, a translated CS curriculum, an interview checklist and a manager's quick reference.
The book and the guides →nginx-lua has 3.5 million Docker Hub pulls and 54 public repositories building on it. The browser-based tools run at kodebeat.com/tools with nothing to install.
Start with a paper
Every tool ships with one: the problem, the method, the numbers, and what the tool deliberately does not do. The book is free; the papers cost an email address.
Read the papers or ask a question