grep for context, 그냥 텍스트.

ctxgrep은 문서, 노트, 기억, 프로젝트 컨텍스트를 검색하는 현지 최초의 CLI입니다. 정확한 검색, regex, semantic retrieval 및 메모리 추출을 결합하여 인간과 AI 에이전트가 현재 작업에 올바른 컨텍스트를 잡아줍니다.

빠른 설정 — ctxgrep에 대한 에이전트를 알려

아래 블록을 복사하고 Claude Code, Cursor, Windsurf 또는 유사한 에이전트 하네스에 붙여 넣습니다. 그것은 ctxgrep을 설치하고, 기술 참조를 읽고, 모든 미래 컨텍스트 검색 작업을 사용하는 방법을 알고.

API 키가 필요하지 않습니다 - ctxgrep은 로컬 embedding 모델로 첫 번째 실행에 자동 다운로드합니다.

특징

• grep-like CLI - 익숙한 명령줄 인터페이스
• Local-first indexing — SQLite + FTS5, 서버 필요 없음 -Hybrid retrieval — 정확하고/ regex/ semantic/ 하이브리드 검색 모드
• Memory Extract - 자동으로 의사 결정, 선호도, 텍스트의 제약을 추출
• Context Packing - 에이전트에 대한 토큰 인용 컨텍스트 번들을 조립
• **JSON 출력 ** - 자동화 및 AI 에이전트에 대한 구조화 출력
• Incremental indexing - 파일과 업데이트만 변경

ctxgrep

grep for context, not just text.

ctxgrep is a local-first CLI for searching documents, notes, memories, and project context. It combines exact search, regex, semantic retrieval, and memory extraction to help humans and AI agents pull the right context into the current task.

Quick setup — tell your agent about ctxgrep

Copy the block below and paste it into your Claude Code, Cursor, Windsurf, or similar agent harness. It will install ctxgrep, read the skill reference, and know how to use it for all future context retrieval tasks.

I want you to use ctxgrep for local semantic search, hybrid retrieval, and
memory extraction across my documents and notes. Please do the following:

1. Install ctxgrep (if not already installed):
   curl -fsSL https://raw.githubusercontent.com/yetone/ctxgrep/main/install.sh | bash

   Or from source:
   cargo install --git https://github.com/yetone/ctxgrep

2. Read the ctxgrep skill reference so you know every command, flag, and
   workflow pattern:
   https://raw.githubusercontent.com/yetone/ctxgrep/main/skills/ctxgrep/SKILL.md

From now on, when you need to search documents, recall past decisions, or
assemble context for a task, use ctxgrep instead of manually grepping files.
For example:
  ctxgrep search --json "authentication design"
  ctxgrep memory --json --type decision "database schema"
  ctxgrep pack --json --budget 5000 "prepare context for the auth refactor"

No API key needed — ctxgrep ships with a local embedding model that auto-downloads on first run.

Features

• grep-like CLI — familiar command-line interface
• Local-first indexing — SQLite + FTS5, no server needed
• Hybrid retrieval — exact / regex / semantic / hybrid search modes
• Memory extraction — automatically extracts decisions, preferences, constraints from text
• Context packing — assembles a token-budgeted context bundle for agents
• JSON output — structured output for automation and AI agents
• Incremental indexing — watches files and updates only what changed
• Provenance — every result includes file path, line numbers, score, and why it matched

Installation

From binary releases

curl -fsSL https://raw.githubusercontent.com/yetone/ctxgrep/main/install.sh | bash

From source

cargo install --git https://github.com/yetone/ctxgrep

Build from source

git clone https://github.com/yetone/ctxgrep
cd ctxgrep
cargo build --release
# Binary at target/release/ctxgrep

Quick Start

# Index your documents
ctxgrep index ~/notes ~/docs ~/meetings --recursive

# Search in the current directory (default — like grep, scoped to CWD)
ctxgrep search "product positioning history"

# Search in specific directories
ctxgrep search "product positioning history" ~/notes ~/docs

# Search the entire index regardless of location
ctxgrep search --global "product positioning history"

# Exact search
ctxgrep search --exact "Lossless Feedback Loop"