eliniscan

eliniscan

**AI-powered full codebase scanner for [Claude Code](https://code.claude.com).** Opens a separate Claude session for every file — reads every line, misses nothing.

I had 942 files and 189,000 lines of code. No matter what I tried — agents, GSD workflows, manual prompting — Claude always skipped lines, summarized instead of reading, and produced shallow reports. After 10+ failed sessions, I built eliniscan. It opened 942 separate Claude sessions, one per file, each with a fresh 1M context window. 11 hours later: **3,894 real issues found**. Every line was actually read.

The Problem

AI code review tools scan diffs or use regex patterns. Claude Code reads your code in conversation, but:

undefined

The Solution

eliniscan takes a different approach:

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│   File 1    │     │   File 2    │     │   File N    │
│ Fresh 1M    │     │ Fresh 1M    │     │ Fresh 1M    │
│ context     │     │ context     │     │ context     │
│ 100% focus  │     │ 100% focus  │     │ 100% focus  │
└──────┬──────┘     └──────┬──────┘     └──────┬──────┘
       │                   │                   │
       └───────────┬───────┘───────────────────┘
                   │
          ┌────────▼────────┐
          │  FINDINGS.md    │
          │  #001, #002...  │
          │  Every issue    │
          │  with line no.  │
          └─────────────────┘

Each file gets its own Claude session. No shared context, no compression, no skipping. The only way to guarantee every line is read.

Install

npm install -g eliniscan

Requires [Claude Code](https://code.claude.com) CLI (`claude` command).

Usage

Inside Claude Code, use these commands:

Scan

/eliniscan:scan

Asks you:

...