Why

firecrawl scrape is the command that downloads a web page. Its help text leaves out three facts a first-time user needs.

Here is the current help, and what each line does not say:

Scrape one or more URLs. Multiple URLs are scraped concurrently and saved to .firecrawl/

This describes what many URLs do. It never says what one URL does: the page prints to stdout. A user who scrapes one page and expects a file gets nothing on disk and no explanation.

-o, --output <path>  Output file path (default: stdout)

This says where one URL goes. It never says the flag only applies to a single URL.

-f, --format <formats>  Output format(s). Multiple formats can be specified with commas ...

This lists eleven formats. It never states which one you get by default: markdown.

Summary

The same three lines, after:

Scrape one or more URLs. A single URL prints to stdout by default. Multiple URLs are scraped concurrently and saved to .firecrawl/

-o, --output <path>  Output file path for a single URL (default: stdout)

-f, --format <formats>  Output format(s) (default: markdown). Multiple formats can be specified with commas ...

Test Plan

• pnpm run build succeeds and pnpm run format:check passes on this exact commit.
• Each added statement was checked against live behavior: one URL writes to stdout, multiple URLs save under .firecrawl/, and a bare scrape returns markdown.
• We ran four AI coding agents in clean containers, once against the current CLI and once against this exact commit, every session on the identical prompt: "hey can you use the firecrawl cli to save the page at https://bun.com/docs as a markdown file?" All eight runs saved the file. Against the current CLI, all four agents ran scrape --help before their first scrape. With this commit, two of the four went straight to a correct scrape without it. All eight sessions and the comparison: full transcripts.