Expert Content Writer
CLI Guide

Two options — standalone binary (no Python required) or pip install.

Option A — Standalone Binary

chmod +x ecw-1.6.0-linux-x86_64
sudo mv ecw-1.6.0-linux-x86_64 /usr/local/bin/ecw
ecw --version

chmod +x ecw-1.6.0-macos-arm64
sudo mv ecw-1.6.0-macos-arm64 /usr/local/bin/ecw
ecw --version

# Rename to ecw.exe and place on your PATH
ecw --version

Option B — Python (3.10+)

pip install .
ecw --version

The pip install gives you the full source — extend, modify, and contribute.

After purchasing on Gumroad you'll receive a license key in your receipt email. Activate it once:

ecw activate YOUR-LICENSE-KEY

Your license is cached locally at ~/.config/ecw/license.json and re-verified every 7 days. If you go offline, a 30-day grace period keeps you writing.

Check your license status anytime:

ecw license-status

Only write and preset-build require a license. Commands like presets, formats, and --help are always available.

ECW calls LLM APIs directly with your own key — no middleman, no markup.

export ANTHROPIC_API_KEY=sk-ant-your-key-here

# Add to ~/.bashrc or ~/.zshrc to persist across sessions

Or use the generic ECW_API_KEY env var, which works across all providers. See Multi-Model Support for OpenAI, Gemini, and Ollama setup.

Write your first article in one command:

ecw write \
  --topic "Why Most Content Marketing Fails" \
  --preset "Professional Blog" \
  --format blog-post \
  --words 1800 \
  --keywords "content marketing, AI writing"

ecw write \
  --topic "Kubernetes Security Best Practices" \
  --urls https://example.com/article-1 https://example.com/article-2 \
  --format technical-brief

The pipeline runs 5 phases automatically: style extraction → skeleton → fat draft → fact-check → executive audit. A final mechanical verification runs 13 automated checks (sentence length, AI phrase detection, readability, and more).

Output

Everything lands in content-workspace/:

content-workspace/
├── published/       # Final articles (Markdown)
├── drafts/          # Intermediate drafts
├── style-profiles/  # Extracted style JSON
├── fact-checks/     # Fact-check reports
└── runs/            # Run manifests

The Preset Builder reverse-engineers any writing style into a reusable voice preset. Three input methods:

From files on disk

ecw preset-build \
  --name "My Brand Voice" \
  --files ~/writing/sample-1.md ~/writing/sample-2.md

From URLs

ecw preset-build \
  --name "Competitor Style" \
  --urls https://example.com/post-1 https://example.com/post-2

From pasted text

ecw preset-build \
  --name "Email Voice" \
  --text "Paste your writing sample directly here..."

The builder extracts sentence patterns, vocabulary fingerprints, punctuation habits, paragraph rhythms, and dozens of other style signals. The output is a Markdown preset file you can use with --preset on every future write.

5 Built-in Presets

ECW works with four LLM providers. Anthropic is the default and recommended — it produces the best results. But you can switch freely.

Provider Setup

pip install expert-content-writer[openai]
export ECW_PROVIDER=openai
export OPENAI_API_KEY=sk-your-key-here

pip install expert-content-writer[gemini]
export ECW_PROVIDER=gemini
export GEMINI_API_KEY=your-key-here

pip install expert-content-writer[ollama]
ollama pull mistral:7b && ollama pull llama3:70b
export ECW_PROVIDER=ollama

Override per command with --provider openai. Override models with ECW_FAST_MODEL and ECW_CAPABLE_MODEL.

Eight format profiles control structure, tone, and length:

Run ecw formats to see word count ranges and structural details for each format.

You pay Anthropic (or your chosen provider) directly. No markup, no per-article fee, no subscription. The tool itself is a one-time purchase.

Cost is based on Anthropic's Claude pricing for a ~2,000 word article. OpenAI and Gemini costs are similar. Ollama runs locally on your hardware — completely free after setup.