Organization repository on GitHub
Semantic code linting with Decision Models
- Forks
- 20
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
Jev directory
CommunityJev Tools & CLIs
Perch
A semantic linting CLI that parses a repository into a method-level call graph, then asks Jev to flag likely defects and check rules written in plain English.
- Stars
- 353
- Language
- JavaScript
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- lakeday-org
- Last verified
- Sep 2026
About this project
perchscan.com · docs · benchmarks · Discord. Semantic code linting with decision models. Run perch login to sign in to Perch Cloud. The first $5 is free. Since decisions are cached, running perch again on unchanged code is free. Perch Cloud also scans pull requests in CI, without writing any workflows. Run perch issues to see them worst first, and perch issues to view one. perch check asks again after you think you've fixed it. It doesn't record anything.
Quoted from the project's README. Read the full README.
Works forJev for Tool calling
Used inJev for Software & DevTools
Highlights
- A sample scan flags wrong_order at 81% confidence and off_by_one at 90%, with severity columns.
- Custom rules live in perch.yaml using a where, each, min, and ensure grammar.
- `perch setup claude-code|codex|pi|cursor` installs agent skills into each tool's directory.
- Dismissed findings are stored in .perch/closed.jsonl and `perch check <id>` re-asks after a fix.
Quickstart
bash
npm install -g @lakeday/perch
export PERCH_API_KEY=<your TypeSafe API key>
perch scanWatch out
Needs a PERCH_API_KEY created in the TypeSafe console, and Perch is a separate product from TypeSafe.
Questions about this entry
Short answers drawn from the linked project or page.
It owns the analysis. perch walks the repository into a method-level call graph and then asks a model about each method with state like the methods it calls and the defects those methods are known for, turning probabilities into a ranked finding list. Any System One endpoint works if you override PERCH_BASE_URL and PERCH_MODEL_ID.
The first $5 on Perch Cloud is free, and decisions are cached, so scanning unchanged code again costs nothing. Signing in also lets Perch Cloud scan pull requests in CI without you writing any workflow files. Bring your own key if you point it at a specific model endpoint instead.
They are what the model returns, surfaced per finding: 81% on a wrong_order defect and 90% on off_by_one in the sample output. The README documents no calibration study, false-positive rate, or precision and recall for those scores. Raw benchmark results are published separately as a Hugging Face dataset.
Severity carries a probability, P1 at 0.8 in the sample output, and perch issues lists findings worst first so you can close what does not matter. Dismissals persist in .perch/closed.jsonl with their reasons, and perch check <id> re-asks about a fix without recording anything.
By default Perch Cloud routes questions to the best model. You can override PERCH_BASE_URL, PERCH_API_KEY, and PERCH_MODEL_ID to use Jev (jev-latest), Liquid AI d1, OpenAI gpt-6-luna, DiffusionGemma Jev, or SemiF; any endpoint speaking the System One format works.
Add them to perch.yaml, or split them across .perch/rules/*.yaml and *.yml and commit them. A rule names a where glob, an each target such as method, and an ensure sentence such as keeping passwords out of logs. perch rules add, edit, and remove manage them.
Perch Cloud can scan pull requests in CI without adding workflows. For agents, perch setup claude-code, codex, pi, or cursor installs the skill or rule file. Findings and dismissed reasons live in .perch/ and .perch/closed.jsonl, which you commit so the ignore list travels with the repo.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Decision outputs in Grep
Hacker News: jevmem keeps Claude Code project memory by asking Jev which turns are worth saving.
avinashjetwani61 pointsJevmem – automatic project memory for Claude Code, built on JevRead the thread on Hacker NewsX: A usage router for Grok Bot
X: The compaction plugin, in Chinese
More Jev resources like this
Rust coding-agent harness (MIT, ~20k stars) whose crates/jcode-base/src/jev.rs is a typed TypeSafe Jev decision transport, used for memory pruning (memory_jev.rs), browser navigation (browser_jev.rs), and voice routing (voice_intent.rs) with separate provider selectors per purpose.
Jev Tools & CLIs#community#rust#coding-agent
Community
1jehuang/jcodeRust
1jehuang/jcodeRustA Rust CLI that ranks agent skills for the next step with a two-stage Jev pass plus lexical prefiltering, with Claude Code hooks, a TUI, abstention, and --why-not explanations.
Jev Tools & CLIs#community#rust#cli
Community
Dicklesworthstone/skillrankerRust
Dicklesworthstone/skillrankerRustAn ast-grep semantic linter where each rule is one natural-language question judged by Jev: 98 rules flag name/body drift, stale comments, weak tests, quiet failures, and commits whose diffs contradict their messages.
Jev Tools & CLIs#community#typescript#cli
Community
mizchi/jev-lintTypeScript
mizchi/jev-lintTypeScriptCLI and GitHub Action that turn project-defined semantic preferences from AGENTS.md into Jev Noul or Choice checks over code changes, then map each result to an advisory or blocking outcome. Checks can be tuned on labeled diffs.
Jev Tools & CLIs#community#javascript#cli
Community
doeixd/jev-prefJavaScript
doeixd/jev-prefJavaScriptA Claude Code plugin and npm library that replaces transcript summarization with Jev keep-or-drop judgments, deleting stale tool calls and results while kept text stays verbatim.
Jev Tools & CLIs#community#typescript#claude-code
Vercel Labs' Unix-like coding agent CLI written in Zig, shipping a typesafe_permission_reviewer builtin that runs the agent's permission decisions through Jev rather than an LLM call. Model-agnostic and embeddable via libfx and ACP.
Jev Tools & CLIs#community#coding-agents#cli
Community
vercel-labs/fxZig
vercel-labs/fxZigFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.