Building tools for structured, reliable AI-driven engineering.
KiniunCorp focuses on how software is designed, built, verified, and shipped in the age of AI — adding structure to AI-assisted workflows, evidence instead of claims, and traceability across the development lifecycle.
Turn AI from a powerful assistant into a dependable engineering system.
Plan→
Execute→
Verify→
Ship
Projects
Three public tools cover three different points where AI-assisted engineering breaks down today — the process around the work, the truth of what the work produced, and the artifact that finally runs. Each is usable on its own.
Spec-To-Ship (s2s)
Open source · MIT
AI chat sessions have no process, memory, or approval gates.
A lightweight process layer around your existing chat client — stages, approvals, and persistent state — while staying fully chat-native and tool-agnostic.
- Routes each request through the minimum stages needed
- Approval gates before any code executes, in an isolated git worktree
- Persistent state and audit trail in
.s2s/ as plain files
- Works with Claude Code, Codex, OpenCode — no lock-in
npm install -g spec-to-ship
s2s init && s2s doctor
bramo-verify
Free · closed engine
Agents report success they never verified.
Audits a git diff and reports what it observed — never what an agent claimed. Your agent says the tests pass; bramo-verify spawns them and reads the real exit code.
- Runs the checks itself, then issues a verdict backed by cited evidence
- A suite that exits 0 having run zero tests is inconclusive, not passed
- Tuned to miss things rather than guess — over-detection is a bug
- The method is published in full: every check, the schema, the limits
npx bramo-verify
CODI
Open source · MIT
Container images ship bloated, unsafe, and unreviewed.
Container Dietitian — a rules-first, AI-assisted container optimisation toolkit. Rules decide; the model only explains and recommends, so results stay reproducible.
- Stack detection and rewrites from a schema-validated rules catalog
- Converts shell-form CMD/ENTRYPOINT to exec form, with rationale
- Security gates and an air-gap guard enforcing zero outbound calls
- Two signed runtimes:
codi:slim and codi:complete
make setup
source .venv/bin/activate
codi run demo/node
Philosophy
Each tool reflects a specific belief about where AI-assisted engineering goes wrong.
- Process over improvisationAI is powerful, but unstructured workflows create risk — so
s2s adds process.
- A claim is not a resultSo
bramo-verify reports only what it observed.
- Determinism beats clevernessWhere correctness matters, CODI lets rules decide and the model explain.
- Simple, transparent, composableEach tool is useful alone and better alongside the others.
Open source, and honesty about it. Spec-To-Ship and CODI are MIT licensed — clone them, read them, fork them. bramo-verify is free to use but its engine is closed-source. We say that plainly rather than letting you discover it, because a tool about honest reporting should be honest about itself.
No core functionality is paywalled. No artificial limits between free and paid. Sponsorship supports development — it does not unlock features.
Support the projects directly →
Working with us
For teams adopting any of our tools — a governed AI workflow with Spec-To-Ship, independent verification of agent output with bramo-verify, or container optimisation with CODI — we may offer setup and enablement guidance, workflow design, and architecture or process reviews. These are scoped independently of open-source sponsorship.
Direction
KiniunCorp is focused on a long-term vision: making AI-native engineering reliable, enabling teams to scale AI workflows safely, and building systems that connect planning, execution, validation, and delivery. More projects will be introduced over time, and existing ones keep moving — bramo-verify's GitHub Action and CODI's real BuildKit builds are both in progress.