fetch

Fetch.Cli — HTML to Markdown

Lightweight CLI for converting web pages to Markdown via Fetch.Server. Output is clean Markdown — ideal for AI agents.

Prerequisites

fetch --version

If not installed:

dotnet tool install --global Fetch.Cli --add-source https://nuget.pkg.github.com/fan92rus/index.json

Requires a running Fetch.Server instance (default: http://localhost:5020).

Basic Usage

fetch "https://example.com"

Conversion Modes

Article mode (default) — extracts main content:

fetch "https://example.com" --mode article

Full-page mode — converts entire page:

fetch "https://example.com" --mode full-page

Short flag:

fetch "https://example.com" -m full-page

Images

Images are excluded by default for minimal token usage. Include them with:

fetch "https://example.com" --images

Set images as default in config:

fetch config set images=true

Loading Types

HttpRequest (default) — fast, no JS rendering:

fetch "https://example.com" --loading-type HttpRequest

Selenium — full browser rendering for JS-heavy pages:

fetch "https://example.com" --loading-type Selenium

Short flag:

fetch "https://example.com" -l Selenium

Custom Server

fetch "https://example.com" --server http://myserver:5020

Short flag:

fetch "https://example.com" -s http://myserver:5020

Configuration File

Set defaults in ~/.config/fetch.cli/config.json:

{
  "Server": "http://localhost:5020",
  "Images": false
}

Priority: CLI flags > config file > built-in defaults.

Interactive setup:

# Via Claude Code slash command
/fetch-init

Manual config:

fetch config set server=http://myserver:5020
fetch config set images=true

View current config:

fetch config

Output Format

Plain Markdown text written to stdout. Suitable for direct use in AI context.

Error Handling

Error	Fix
Server error (502)	Fetch.Server is not running. Start it with dotnet run --project Fetch.Server
Connection refused	Check server URL and network connectivity
Server returned empty response	Page may be empty or blocked. Try --mode full-page
Empty JS-rendered content	Use --loading-type Selenium with a running browser

Best Practices

• Use article mode by default — cleaner, more relevant content
• Exclude images by default — saves tokens for AI context
• Use full-page mode when you need navigation, sidebars, or complete page structure
• Use Selenium only when page content requires JavaScript rendering
• Keep server running — Fetch.Server must be accessible for the CLI to work

CLI Reference

See CLI_REFERENCE.md for complete documentation.