kindintelligence/jev-rust-review
User repository on GitHub
Rust-aware code review for Claude Code and coding agents, powered by TypeSafe Jev
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Agents & MCP
Jev Rust Review
Claude Code plugin for Rust review that runs Clippy and cargo-semver-checks itself, has Jev triage changed units for defects no tool reports, then verifies each candidate finding before it reaches you.
- Stars
- 2
- Language
- Rust
- License
- MIT
- Published by
- Community
- Author
- kindintelligence
- Last verified
- Oct 2026
About this project
A Claude Code plugin for Rust code review that goes deeper than the compiler. rustc, Clippy and cargo-semver-checks report what tools can find, filtered to the lines you changed. TypeSafe's Jev model and Claude spend their effort on what no tool reports: a select! branch that loses data when it is cancelled, a check and an act under two separate locks, a length guard removed from a caller in another file.
Quoted from the project's README. Read the full README.
- Languages
- Rust
Highlights
- One cargo clippy --message-format=json run supplies the facts: errors anywhere, warnings on changed lines only, and Claude never reads raw output.
- Every Jev question carries a sentence saying why no compiler check or lint answers it, and a test fails if one is empty.
- Findings are verified before display as supported, refuted or insufficient_context, plus a severity score and a real-defect versus taste call.
- Lints the project allows via #[allow] or [lints.clippy] stay quiet, and a defect a tool already reported never returns as a flag.
Watch out
It is in active development at v0.1 and the README says interfaces, question ids, thresholds and output formats may change between versions. The first launch builds a Rust binary from source, 65 to 75 s on an 8-core machine against Claude Code's 30 s MCP startup timeout, so the server shows as failed until you reconnect. It needs Claude Code 2.1 or later on Linux or macOS; Windows works through Git Bash and is not covered by CI. cargo-semver-checks is optional and never installed by the plugin, and Miri is suggested rather than run when unsafe code changes. The README lists Jev's documented weak spots as literal reading, counting, indirection, prompt injection and large irrelevant state, and notes that complementary questions are not guaranteed to agree. It is an independent project that TypeSafe and Anthropic do not endorse. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
The ordering is the point. The plugin runs the tools itself and treats their output as fact, then asks Jev only about defects the tools structurally cannot see, such as a select! branch that loses data when cancelled or a check and an act under two separate locks. Where a tool does find the pattern, the question keeps only the judgement, and anything a tool reported on the same lines is dropped from the Jev output.
The launcher tries a cached binary, then one on PATH, then a release binary checked against SHA256SUMS, then falls back to a background cargo build --release. Because that build takes longer than Claude Code's 30 s startup timeout, the server shows as failed with a message pointing at the build log; run /mcp and reconnect jev when it finishes, or run scripts/launch.sh --install in the foreground instead.
Jev input is billed at $0.042 per million tokens, and each changed unit gets up to 42 core questions plus framework questions, sent as one fan-out request per unit, with a second verification pass over findings. Every number in a report is labelled with its source, and evaluate_rust_changes returns token usage and cost for the run. No per-review total is published, so measure it on your own diffs.
The design sets a low bar to flag during triage, since a flag only decides what Claude reads first and a unit with no flag is still reviewed, and a high bar to report afterwards. Verification asks whether the code contains the claimed defect, how severe it is and whether it is real, a matter of taste or too remote to matter. insufficient_context findings survive only when Claude's own confidence is High, and the report says Jev could not verify it.
Claude Code 2.1 or later on Linux or macOS, a TypeSafe API key, git, and cargo with Clippy. cargo-semver-checks is optional and only used when a library's public surface changed.
Yes. Without a key it runs as a Claude-only review under a banner, and the check that stops a defect being reported twice, once by a tool and again by Jev, also runs without a key.
Yes. /jev-rust-review:rust-review --dry-run shows exactly what would be sent and sends nothing, and --json ends the report with a machine-readable list of what it reported.
More Jev resources like this
35GitHub stars
Rust terminal tool that asks Jev noul, choice and score questions about piped text or files and returns calibrated probabilities with stable exit codes for shell scripts, CI jobs and agents.
Jev Agents & MCP#community#jevbest#catalog-import
Community
shaharia-lab/jev-cliRust
shaharia-lab/jev-cliRust5GitHub stars
Grok Build wrapper that asks Jev one noul per tool before the first model call of a turn, then holds the shortened tool list so the prompt cache survives.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
daniel-farina/nitroRust
daniel-farina/nitroRust10GitHub stars
GPL-3.0 desktop app for Linux, macOS and Windows that gives every coding-agent task its own git worktree and its own terminal, with live diffs and optional Jev review of changed files.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
joaoh82/yardsortRust
joaoh82/yardsortRust51GitHub stars
Local LinkedIn recruiting workspace on Browser Use's Jev Ultrafast: Jev screens titles, drives navigation, and assesses profile excerpts against your brief while a live browser stays visible; matches and evidence export as JSON.
Jev Agents & MCP#community#jevbest#catalog-import
Community
skeptrunedev/jev-recruiterPython
skeptrunedev/jev-recruiterPython33GitHub stars
Codex Skill plus stdlib-only Python CLI that recommends which installed skill to read next for a task, allowing no-skill or uncertain answers and never executing candidates.
Jev Agents & MCP#community#jevbest#catalog-import
Community
win4r/jev-skill-suggesterPython
win4r/jev-skill-suggesterPython72GitHub stars
MCP server over stdio that gives coding agents fourteen Jev judgment tools, turning evidence plus an enumerable question into auto, review, or escalate actions from calibrated probabilities.
Jev Agents & MCP#community#jevbest#catalog-import
Community
PyModel/jev-judge-mcpPython
PyModel/jev-judge-mcpPythonFrom the Jev community
Posts from builders shipping with Jev right now.