Kibi is a requirement-conformance layer for AI coding agents that keeps product intent connected to code and tests, and catches when implementations drift from
Joined August 2026
- Tweets121
- Following73
- Followers8
- Likes57
Kibi's numeric checks are Prolog: intervals.pl folds each comparison on a variable into an interval and asks if it's empty.
Now those folds are proved exact in Lean 4, and CI runs 200k cases through the Prolog code, the Lean model and a CLP(Q)/CLP(FD) oracle.
3/ We also stopped calling passing E2E tests "proof". The old "proven" status is now "tested".
No proof kind issues a certificate for code yet, so today nothing reaches "proved". The ladder exists so no guarantee reads stronger than what backs it.
Update: as of today something can. The formal proof lane is out: certificates that cover every case, re-checked by Kibi itself. xitter.cf/kibi_dev/status/211462…
6/ Each answer leaves a certificate: a model when the facts are consistent, an irreducible core when they aren't.
A small separate checker re-checks every certificate by evaluation only. It never reruns the search that found it.
kibi check --write-certificates certs.json
10/ It's plain Markdown, so it renders wherever your agent replies, chat apps and cloud sessions included, plus a self-contained HTML page.
/kibi-signoff in Claude Code, Cursor, ZCode and OpenCode. VS Code opens it in a panel. Hooks ask for it once an agent writes requirements.
11/ Kibi is a repo-local knowledge base for coding agents, checked by Prolog (bundled on Linux and macOS).
npm i -D kibi-core kibi-cli kibi-mcp
github.com/Looted/kibi Proof ladder: looted.github.io/kibi/guide/…
5/ A contradiction comes back as the smallest set of facts that can't hold at once, plus the requirements behind them, so you know what to change or supersede.
Anything it can't compare (mixed units, unsupported operators) is reported as unresolved, never as consistent.
4/ New in kibi check: it proves your recorded facts are consistent. One value satisfies every fact on a property, every allow/deny decision has one outcome, stored rules are stratified.
It also catches facts that contradict only together, which pairwise checks miss.
1/ Coding agents follow the current prompt and quietly drop what was agreed earlier.
Week 1: "sessions expire after 30 min idle."
Week 3: "make sessions last longer." Tests stay green, the rule is gone.
Kibi, an open-source requirements checker for coding agents, catches this.
6/ To try it (Node 22+): copy the setup prompt from the README into your agent.
It installs Kibi in the project, connects itself, asks where your intent lives (tickets, specs, wikis) and drafts a plan citing them. Nothing is written until you approve.
github.com/Looted/kibi#quick…
7/ It's beta and while the mission statement and feature set are mostly stable, we keep improving on accuracy and expanding the functionality. What helps most right now is hearing where it gets in your way, here or in GitHub issues.
Docs: looted.github.io/kibi