Comparisons · Last updated 2026-09-21

Jev vs code analyzers

Code analyzers (linters, typecheckers, SAST, Semgrep) prove or pattern-match on an AST. Jev can help an agent pick a skill or decide “does this diff look like a secrets leak in prose?” It will not compile your repo.

Unofficial. TypeSafe’s agent skill teaches coding agents to batch System One questions. That does not replace tsc or CodeQL. docs.typesafe.ai.

Comparison scope

Use analyzers for definite defects. Use Jev for review routing, “needs human,” or catalog skill suggestion (their cookbook) — not for numeric complexity scores.

Criteria that decide the architecture

Decision quality and control

Official jaggedness: high-level language questions beat assembly/binary. Do not send minified bundles as state and expect a security verdict.

Integration trade-offs

CI: analyzers block merge on definite issues. Optional Jev on the PR description + selected hunks for “needs security review.” Pin the model. Keep diffs small.

TypeSafe’s public models page lists jev-1.13 at $0.042 per million input tokens with output tokens free — a vendor claim, not a jev.pro measurement. Confirm on the models page before you bid.

When each approach fits

Prefer Jev when

• You need a snap judgment on review comments or routing

Prefer Code analyzers when

• The question has a formal answer in the AST

Analyzers first; Jev routes leftover judgment; humans merge.

Route the leftover comment

Linters, typecheckers, SAST, and Semgrep prove or pattern-match on an AST. Jev will not compile your repo. Sketch: CI analyzers block merge on definite issues → optional Jev on the PR description plus selected hunks → Choice/Score “needs security review” / “docs-only” / other on pinned jev-1.13.0 → route the review. Keep diffs small. Official jaggedness: high-level language questions beat assembly/binary. Do not send minified bundles as state and expect a security verdict.

TypeSafe’s agent skill teaches coding agents to batch System One questions (claude plugin / npx skills add typesafe-ai/skills). That is a drop-in for the agent, not a replacement for tsc or CodeQL. Read SKILL.md upstream; do not dump it here. Review-comment hops use the vendor-claimed $0.042 / M input list price — still not a SAST license. Jev does not generate the fix. The coding agent writes; Jev may gate.

How a “Jev linter” ships a type error

• Numeric complexity Score. Jaggedness: do not reconstruct magnitudes. Cyclomatic counts stay in the analyzer.
• Secret regex as a second-class hop. Analyzers win. Jev on prose (“does this comment look like a leaked token story?”) is leftover, not the first line.
• Whole minified vendor bundle. State rot plus encodings Jev is weak on. Specialized tools for hex/binary.
• Skill-suggestion cookbook as SAST. Official: rank skills, Noul whether any is needed. Still not a linter.
• Schema-safe “looks safe” as a merge. Valid output is not a clean CodeQL run. Humans merge.

When not to use Jev here

Do not use Jev when the question has a formal answer in the AST, when you need secret regex or type errors, or when the artifact is binary. Prefer analyzers first. Use Jev for a snap on review comments or routing leftover judgment.

What this page does not claim

• Not a SAST vendor chart.

• Schema-safe output is not the same as factual correctness.

• No independent bake-off numbers live here.

FAQ

Can Jev write the fix? No generation. The coding agent writes; Jev may gate.

Skill suggestion cookbook? Official: rank skills, Noul whether any is needed. Still not a linter.

Does the agent skill replace Semgrep? No. It teaches agents to batch Jev questions. Analyzers still own AST truth.

Disclaimer

This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.

Hub: Comparisons. Siblings: vs security automation, agent skill, questions for builders. Canonical: https://docs.typesafe.ai.