On-device checks in Terminal

AI Content Detector CLI for Claude, Codex, and Cursor

The slop CLI runs Slop or Not's AI text detector and AI image detector from your Mac terminal. It returns JSON for Claude, Codex, Hermes Agent, OpenClaw, Cursor, scripts, and editors, and it also covers raw image-model scoring, readability, and text cleanup.

Moving from a cloud detector like ZeroGPT or GPTZero? Pipe in a draft, a screenshot, or a local file. Slop or Not scores it locally and never uploads it.

Runs locally: Text, image, readability, and cleanup checks stay on your Mac.
Outputs JSON: Use --json when scripts or agents need stable results.
Unlimited local checks: The CLI ships with Pro. No daily check cap, no per-call metering, so scripts and agents can rerun text, image, readability, and cleanup checks as often as they need.

Setup

How do I install the Slop or Not CLI?

Install Slop or Not for Mac, open it once, and activate Pro. Then run the setup block below: it links the bundled binary into ~/.local/bin and puts that folder on your PATH, which macOS does not do for you. Settings > Command Line offers a different one-liner that puts the app bundle itself on your PATH. Verify with slop status. From here, every new Terminal session finds slop on PATH with no reinstall.

# One-time setup for zsh, the default shell on macOS.
mkdir -p ~/.local/bin && ln -sf '/Applications/Slop Or Not.app/Contents/MacOS/slop' ~/.local/bin/slop
grep -qsxF 'export PATH="$HOME/.local/bin:$PATH"' ~/.zshrc || printf '\nexport PATH="$HOME/.local/bin:$PATH"\n' >> ~/.zshrc
source ~/.zshrc

slop status --json

Command reference

What can the Slop or Not CLI do?

The command surface is small on purpose: status, text, image, score-image, readability, cleanup, and MCP.

Every example below is real output, captured from Slop or Not 1.1.1 for Mac in August 2026 and abridged where marked.

slop status

Check your install

Confirms the CLI can find the app bundle and reports whether Pro is active.

View command and response

Run

slop status --json

Returns

{
  "localCachePro" : true,
  "pro" : true,
  "version" : "1.1.1"
}

slop text

Score AI text

Reads UTF-8 text from stdin and returns the same on-device AI probability the Mac app shows, plus readability metrics. Add --language when you already know the language.

View command and response

Run

pbpaste | slop text --json

Returns

{
  "detectedLanguage" : "en",
  "detection" : {
    "result" : {
      "_0" : 0.12590113,
      "_1" : { "most_likely_real" : {} }
    }
  },
  "inputTruncated" : false,
  "readability" : {
    "detectedLanguage" : "en",
    "languageConfidence" : 0.99984,
    "scores" : [
      { "kind" : "fleschReadingEase", "direction" : "higherIsEasier", "label" : "Fairly difficult", "value" : 50.85 },
      { "kind" : "fleschKincaidGradeLevel", "direction" : "gradeLevel", "value" : 16.97 }
    ],
    "full" : { "consensus" : { … }, "genre" : { … }, "gradeLabel" : { … }, … },
    "selectedLanguage" : "en",
    "stats" : { "wordCount" : 87, "sentenceCount" : 2, "syllableCount" : 115, … },
    "warnings" : [ … ]
  },
  "sentenceCount" : 2
}

slop readability

Analyze readability

English gets Flesch Reading Ease and Flesch-Kincaid Grade Level, plus Gunning Fog, Coleman-Liau, and ARI in the full report, with a consensus grade that discards outliers. German, Spanish, Italian, Swedish, Danish, and Norwegian each use their established formula.

View command and response

Run

cat draft.md | slop readability --json

Returns

{
  "readability" : {
    "detectedLanguage" : "en",
    "languageConfidence" : 0.99984,
    "scores" : [
      { "kind" : "fleschReadingEase", "direction" : "higherIsEasier", "label" : "Fairly difficult", "value" : 50.85 },
      { "kind" : "fleschKincaidGradeLevel", "direction" : "gradeLevel", "value" : 16.97 }
    ],
    "full" : { "consensus" : { … }, "genre" : { … }, "gradeLabel" : { … }, … },
    "selectedLanguage" : "en",
    "stats" : { "wordCount" : 87, "sentenceCount" : 2, "syllableCount" : 115, … },
    "warnings" : [ … ]
  }
}

slop cleanup

Clean hidden text tricks

A lookup table, not a rewrite: strips zero-width characters, homoglyphs, and fancy punctuation without touching your words. Add --british to convert American spellings to British English.

View command and response

Run

pbpaste | slop cleanup --json

Returns

{
  "britishMappings" : [],
  "cleanedText" : "He said \"hello\" - and left.",
  "detectedLanguage" : "en",
  "homoglyphCounts" : [],
  "invisibleCounts" : [
    { "count" : 1, "kind" : "Zero Width Space (U+200B)" }
  ],
  "punctuationCounts" : [
    { "count" : 1, "kind" : "Em Dash (U+2014) with '-'" },
    { "count" : 1, "kind" : "Left Double Quotation Mark (U+201C) with '\"'" },
    { "count" : 1, "kind" : "Right Double Quotation Mark (U+201D) with '\"'" }
  ]
}

slop image

Detect AI images

Checks piped JPEG, PNG, HEIC, or WebP image bytes locally. Add --ocr to read the text inside a screenshot.

View command and response

Run

slop image --json < ~/Desktop/screenshot.png

Returns

{
  "detection" : {
    "result" : {
      "_0" : 0.8416403,
      "_1" : { "most_likely_ai_slop" : {} }
    }
  },
  "detectionSourceRaw" : "ml"
}

slop score-image

Score images with the image model

Pass a piped image straight to the model and get its raw score, skipping the provenance heuristics. Useful when an image-generation loop needs direct model feedback. Requires the AI image detection model to be installed.

View command and response

Run

slop score-image --json < ~/Desktop/generated.png

Returns

{
  "rawSlopScore" : 0.706787109375
}

slop mcp

Start the MCP server

Runs the stdio server that exposes all six tools to Claude, Codex, Cursor, and other MCP clients. Use --print-config to print ready-to-paste setup snippets instead of starting the server.

View command and response

Run

slop mcp --print-config

Returns

CLIENT SETUP
------------
To use slop CLI in the Terminal, run this once to symlink it
into ~/.local/bin:

  mkdir -p ~/.local/bin && ln -sf '/Applications/Slop Or Not.app/Contents/MacOS/slop' ~/.local/bin/slop

MCP SERVER
----------
Run slop CLI as a MCP server so your AI client can call its
detection tools directly. …
(snippets for Claude Desktop, Claude Code, Codex CLI, and Cursor follow)

Fast paths

Common terminal workflows

These are the commands most people reach for first. Each one reads local input and prints a local result.

How do I score AI text from Terminal?

Copy a draft, pipe it into slop text, and read the JSON verdict. Add --json when another tool needs stable output.

pbpaste | slop text --json

How do I check readability?

Readability is separate from AI detection. Use it when you need a reading-level check without sending text to a web tool.

cat draft.md | slop readability --json

How do I clean hidden AI fingerprints?

Text Cleanup strips invisible characters, homoglyphs, and spacing tricks. It does not rewrite the text.

pbpaste | slop cleanup --json

How do I get a raw image-model score?

Use score-image when an image-generation workflow needs the raw model signal, not the provenance-aware image detection result.

slop score-image --json < ~/Desktop/generated.png

How do I check an image file?

Pipe in screenshots, product shots, dating profile photos, and other files you want checked locally.

slop image --json < ~/Desktop/screenshot.png

Local AI detector API

Can I use the CLI as a local AI detector API?

For local automation, yes. The CLI reads stdin or local files and returns JSON, so scripts and editors can treat it like a local AI detector API without uploading text or images to a cloud endpoint. It is not a hosted HTTP API; the Mac app runs the check on the machine that calls the command.

pbpaste | slop text --json
slop image --json < ~/Desktop/source-photo.png
slop score-image --json < ~/Desktop/generated.png

MCP from CLI

How do I start MCP from the CLI?

The MCP server is a CLI subcommand. Run it directly when your MCP client asks for a command, or when you are debugging setup.

slop mcp

Set up MCP for Claude, Codex, Hermes Agent, OpenClaw, and Cursor Read the launch post agentic AI Humanizer guide