When should I use this instead of plain text search?

Use text search when you know the symbol or literal. JevGrep is for when you only know the behavior, such as where session expiry is handled, across implementation and tests.

What source code actually leaves my computer during a search?

Eligible source fragments go to the configured provider with your question, paths, line ranges, and the relevance criterion. Inspect and doctor commands stay offline for pre-checks.

How do credentials and per-repository authorization work?

Keys can be stored per machine or passed as environment variables, while each repository has its own trusted profile stored outside the repo with an explicit remote-evaluation switch.

Should I run it as a CLI or as an MCP server for my agent?

Run it as a CLI for terminal search or as a stdio MCP server exposing one semantic_search_code tool for agents, with one server process per repository.

Which providers can JevGrep score with?

TypeSafe AI with jev-1.13.0 pinned, Vercel AI Gateway with typesafe-ai/jev, or OpenRouter with typesafe/jev-1.13. TypeSafe is proposed first during jevgrep init --global.

How do I check scope and limits before sending code?

Run jevgrep doctor to check provider, credential, authorized root, limits, and cache without a network request, and jevgrep inspect to see which files and fragments are eligible and how much work a search would do.

What tool does the JevGrep MCP server expose?

One stdio tool, semantic_search_code. Starting the server does not scan files or contact a provider; a tool call performs a search using the authorization for the current directory, and one server process serves one repository.

CommunityJev Agents & MCP

Jevgrep — nassim-arifette

Behavioral code search that asks Jev to score repository fragments for a plain-language question and returns exact excerpts with paths and lines.

See highlights

Open github.com

Stars: 102
Language: TypeScript
License: MIT
Published by: Community
Author: nassim-arifette
Last verified: Oct 2026

nassim-arifette/jevgrep

User repository on GitHub

View on GitHub

Jev-powered semantic code search for coding agents — find behavior across repositories via CLI or MCP, with exact source excerpts and line numbers.

Forks: 11
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Find code by what it does. JevGrep helps coding agents find relevant code when they do not know the file name or symbol to search for. Ask a question such as “Where is session expiry handled?” and JevGrep scans the authorized repository, asks Jev to score all eligible fragments, then returns the original source excerpts with their paths and line numbers. The calling agent can read those files in detail and continue its work with less exploratory context.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import

Works forJev for Classification Jev for Ranking Jev for Tool calling

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Plain-behavior queries return original source excerpts with paths and line numbers instead of generated summaries.
Global provider setup for TypeSafe, Vercel AI Gateway, or OpenRouter with per-repository authorization and ignore files.
Offline doctor and inspect commands check credentials, scope, exclusions, and expected work before sending code.
CLI plus stdio MCP server with one search tool and documented setup for common coding agents.

Watch out

Each repository must be authorized separately and remote evaluation explicitly enabled; excluded files and scan caps control what leaves the machine. 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.

Other listings of this project

Forks and copies of this project in the directory. This entry is the canonical one.

jevgrep

A streaming grep-by-meaning CLI that judges each line with one Jev yes/no question and prints the lines above a threshold, built to sit behind tail -f, kubectl logs -f, or journalctl -f.

Jev Tools & CLIs#community#python#cli

Community

allebee/jevgrepPython

More Jev resources like this

pi-jev-skill-picker

Pi extension that strips the full Agent Skills catalog from every request and loads only the skills a Jev ranking call judges relevant.