curl -O / curl -o file: download a file with curl

curl -O https://example.com/file.zip downloads the file and saves it in the current directory under the name from the URL (file.zip). curl -o name.zip URL saves it under a name you choose. Add -L to follow redirects, which most download links use, and -f so that an HTTP error such as 404 fails with a non-zero exit status instead of saving the error page. The usual safe form is curl -fLO URL.

Example

A small Node.js server on 127.0.0.1 stands in for a download site; it serves one file, sends a Content-Disposition file name, answers 404 for anything else, and stops when the block requests /quit:

cat > server.js <<'EOF'
const s = require("http").createServer((req, res) => {
  if (req.url === "/quit") { res.end(); return process.exit(0); }
  if (req.url === "/files/report.csv") {
    res.writeHead(200, { "Content-Type": "text/csv", "Content-Disposition": 'attachment; filename="q3.csv"' });
    return res.end("id,total\n1,10\n");
  }
  res.writeHead(404, { "Content-Type": "text/plain" });
  res.end("not found\n");
});
s.listen(5140, "127.0.0.1");
setTimeout(() => process.exit(0), 20000);
EOF
node server.js &
sleep 1
U=http://127.0.0.1:5140
curl -s -O "$U/files/report.csv"
curl -s -o data.csv "$U/files/report.csv"
curl -s -OJ "$U/files/report.csv"
curl -s --create-dirs -o downloads/report.csv "$U/files/report.csv"
ls *.csv downloads
cat data.csv
curl -s -O "$U/missing.zip"; echo "without -f: exit $?, saved: $(cat missing.zip)"
curl -sf -o other.zip "$U/missing.zip"; echo "with -f: exit $?"
curl -s "$U/quit"

data.csv
q3.csv
report.csv

downloads:
report.csv
id,total
1,10
without -f: exit 0, saved: not found
with -f: exit 22

-O used the last part of the URL, report.csv. -o chose the name. -J together with -O took the name from the Content-Disposition header, q3.csv. --create-dirs created downloads/. The last two commands show why -f matters: without it the 404 page is saved as missing.zip and curl exits 0; with it curl exits 22 and saves nothing.

Options

Pitfalls

• URLs with query strings. With -O, https://host/get?id=7 is saved as get (curl 8.21 drops the query part); use -o to choose a proper name.
• Unquoted URLs. &, ? and [ have meanings in the shell and in curl's URL globbing. Quote the URL, and use -g if it contains literal brackets.
• Overwriting. -o and -O overwrite an existing file without asking. Only -J refuses to overwrite: curl 8.21 prints Warning: Failed to open the file q3.csv: File exists and exits with code 23.
• Windows. In PowerShell 5.1 curl is an alias of Invoke-WebRequest; call curl.exe to get real curl, which ships with Windows 10 and later.

Confirm the download

Check the exit status (echo $?, 0 means success), then the file: ls -l file.zip, file file.zip, and if the site publishes a checksum, sha256sum file.zip. -w '%{http_code} %{size_download}\n' prints the status code and the number of bytes received after the transfer.

FAQ