Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

klint

Semantic linter that evaluates files against plain-English JSON rule packs in one /v1/systemone pass, locally with Kev or remotely with Jev.

Modelbeta
View on GitHub ↗

Overview

klint aims to sit between syntax-bound linters such as ESLint or Ruff and slow, expensive LLM code reviewers. Teams define architectural rules or antipatterns in plain English in an ESLint-style klint.json file, and klint evaluates files and package trees against a whole rule pack in a single forward pass with zero decoded tokens using the System One pointer-head interface.

It can run entirely offline with open-weight Kev models (0.8B or 4B) from Hugging Face, taking about 0.5 to 2.5 seconds per file on Apple Silicon or less on CUDA, or point to any remote /v1/systemone endpoint such as hosted Jev or a self-hosted Kev server. Its hexagonal architecture uses auto-registering auditable units, pluggable extractors, rule loaders and model evaluators so it can be adapted to other languages. The README positions it against deterministic linters, generative PR bots and other approaches in a comparison table. It suits teams wanting to codify design rules that syntax cannot express.

What the decision model decides

TypeQuestionAnswers
Yes / noDoes this file or unit violate this plain-English rule?(inferred)—

How it works

  1. Write rules in klint.json
  2. klint extracts auditable units from files
  3. All rules in a pack are evaluated in one forward pass
  4. Violations above threshold are reported

Strengths

  • Can run fully offline with open weights
  • One pass for a whole rule pack
  • Pluggable architecture for other languages

Caveats

  • Local inference speed depends on hardware
  • Accuracy depends on small model quality and rule wording
  • Early project

At a glance

Type
Model
Status
beta
Language
Python
Runtime
CLI with local Kev or remote endpoint
Has code
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

Does it need a cloud API?

No, it can run offline with Kev models; remote Jev is optional.

How are rules written?

In plain English in an ESLint-style klint.json file.

How fast is it locally?

About 0.5 to 2.5 seconds per file on Apple Silicon, per the README.

Related projects

  • JevGate

    Rust code-review gate asking Jev small typed questions about functions and tests, failing CI only on rules measured 80% right.

    CLIproductionRust
  • CodeOtter★ 3

    Self-hosted PR review platform pairing an LLM narrative with a /v1/systemone decision engine for 6 rubric scores and 8 yes/no merge gates.

    Servicebeta
  • Abide

    Coding-agent hook that asks Jev one yes/no per project rule on every edit or turn and makes the agent fix any violation.

    HookbetaTypeScript
  • code-quality

    Advisor for coding agents: git and Stop hooks flag test tampering, secrets and skipped hooks, with optional Jev review findings.

    Hookbeta
  • jeff (Alurith)

    Read-only Go CLI that checks files against your plain-language code rules with Jev, locally or in CI.

    CLIbetaGo
  • jev-codes

    CLI and agent commands auditing a git diff against YAML standards packs, scoring each hunk with Jev so agents fix only flags.

    Agent skillbetaTypeScript