Beautiful Soup is a Python library that turns HTML into a tree you can search, first released in April 2004 and now at version 4.15.0. Its history is one author, 85 releases and 22 years. In all that time it has never fetched a page, which is why the proxy goes somewhere else.
A tool's history is usually read from its commit log. Beautiful Soup's code host tells automated readers to stay out, so we left that log closed. On November 11, 2026, we downloaded every release file from the author's own archive, 101 files. We read the full changelog and compared the code of the releases with each other. We did not run any of it.
What is Beautiful Soup in version 4.15.0?
Version 4.15.0 is dated June 7, 2026. You hand it a document and the name of a parser. It gives back a tree of Python objects.
- It has no parser of its own. Three parsers plug in: Python's built-in
html.parser, lxml and html5lib. - Search.
find()andfind_all()look for tags by name, attribute, text or a function of your own.select()takes CSS selectors through the SoupSieve package. - Encodings. A class named
UnicodeDammitworks out the encoding and turns the document into Unicode. - Changes. Methods insert, replace, wrap and remove elements, and
prettify()writes the tree back out. - Type hints cover all public classes and methods since version 4.13.0.
The library holds 10,649 lines of Python in 15 files, with 553 test functions. It needs Python 3.7 or newer and carries the MIT license. It requires two other packages, soupsieve and typing-extensions. PyPI Stats counted 289,647,647 downloads of it in the month before November 11, 2026.
Do I install bs4, beautifulsoup4 or BeautifulSoup?
pip install beautifulsoup4 lxml
Three package names exist on PyPI, and only one is the library.
- beautifulsoup4 is the real one. Since 2012 you import it as
bs4. - bs4 is, in the words of its own page, "a dummy package designed to prevent namesquatting". It was still downloaded 15,850,654 times in a month.
- BeautifulSoup is version 3. It was discontinued on January 1, 2021, and was still downloaded 165,639 times in a month.
Then name the parser, as in BeautifulSoup(html, "lxml"). Without a name, the documentation says, "you'll get the best HTML parser that's installed. Beautiful Soup ranks lxml's parser as being the best, then html5lib's, then Python's built-in parser." The same script can then build two different trees on two machines. Our BeautifulSoup tutorial shows how far the three parsers differ.
Who builds Beautiful Soup?
Leonard Richardson started it in 2004. The AUTHORS file of 4.15.0 still says he "is the primary maintainer", and PyPI lists one maintainer. The file opens with "Behold, mortal, the origins of Beautiful Soup..." and names three people with significant contributions: Aaron DeVore, Isaac Muse and Chris Papademetrious. It also records a prize: "Jonathan Ellis was awarded the prestigious Beau Potage D'Or for his work in solving the nestable tags conundrum."
Money was never the point. On a page of March 2009 Richardson wrote: "To justify further Beautiful Soup development I need time, not money." Today the home page points companies to a Tidelift subscription.
In 2017 he wrote a short zine named Tool Safety about what the library taught him. It says the library has "also been used as a tool for spamming, doxing, and other things I think are unethical". It also says: "If I'm going to feel pride about the good people do with my software, I have to share some responsibility for the bad." He stopped asking for donations. The best thanks, he writes, "is to be thoughtful about how you use it".
How did Beautiful Soup change in 22 years?
Apr 2004
Version 1.0
one file on SGMLParser
May 2006
Version 3.0.0
Unicode and findAll
Dec 2008
Version 3.1.0
moved to HTMLParser
Nov 2010
Version 3.2.0
back on SGMLParser
Mar 2012
Version 4.0.1
three parsers plug in
Dec 2018
Version 4.7.0
SoupSieve for CSS
Sep 2021
Version 4.10.0
Python 3 only
Feb 2025
Version 4.13.0
type hints
Jun 2026
Version 4.15.0
last with the old names
2004: one file and a song
The changelog dates version 1.0 to April 20, 2004, with two words: "Initial release." Its name is "So rich and green".
The names come from a song. In Chapter X of Alice's Adventures in Wonderland, the Mock Turtle sings "Beautiful Soup, so rich and green, / Waiting in a hot tureen!" The releases follow it line by line. Version 1.1 is "Swimming in a hot tureen", where the song says Waiting. Version 1.2 is "Who for such dainties would not stoop?" Version 2.0.0 is "Who cares for fish?", and 3.0.0 is "Who would not give all else for two p", cut in the middle of a word as the book prints it. After 3.0.0 the song stops.
The oldest file in the archive holds 501 lines, five classes and three imports. Its header says: "This library has no external dependencies. It works with Python 1.5.2 and up." It sat on SGMLParser, a parser from Python's standard library.
2005 and 2006: two rewrites
Version 2.0.0 of April 10, 2005 opens with a verdict on its past: "Beautiful Soup version 1 was very useful but also pretty stupid." It added a class for odd pages named ICantBelieveItsBeautifulSoup, which stayed in the code until the last release of version 3.
Version 3.0.0 of May 28, 2006 broke old code on purpose. Every document became Unicode, SoupStrainer could parse part of a page, and the search methods got the names find and findAll, taken from Rubyful Soup. Five days later, version 3.0.2 brought the older names back as aliases, with a promise: "Those names are deprecated and if I ever do a 4.0 I will remove them. I will, I tell you!"
2008 to 2010: the release to forget
Python 3.0 removed SGMLParser. Version 3.1.0, whose files are dated December 29, 2008, moved to HTMLParser so the code could run on Python 3. The author's page about it says the new parser "did significantly worse on real-world HTML".
His advice was unusual: "You can pretend that Beautiful Soup version 3.1.0 was never released." The same page says: "I no longer enjoy working on Beautiful Soup, but too many people depend on it for me to let the project die". He went back to the old parser with version 3.0.8 in December 2009. In November 2010 the changelog closed the matter: "The 3.1 series wasn't very useful, so I renamed the 3.0 series to 3.2 to make it obvious which one you should use."
2012: Beautiful Soup 4
The answer was to stop parsing. The beta of February 3, 2012 calls version 4 "a nearly-complete rewrite that removes Beautiful Soup's custom HTML parser in favor of a system that lets you write a little glue code and plug in any HTML or XML parser you want." The module got the name bs4.
The changelog lists eight betas in 28 days. Version 4.0.1 of March 14, 2012 is the first official release, and the changelog explains the number: "There is no 4.0.0 release, to eliminate any possibility that packaging software might treat "4.0.0" as being an earlier version than "4.0.0b10"." The year holds 18 releases. One file of 2,015 lines became eight files of 2,426.
2013 to 2021: quiet years, and Python 2 leaves
The longest silence lasted 639 days, from October 2013 to version 4.4.0 of July 3, 2015. That release added the warning for a missing parser name, select_one() and a standard MIT license. It also renamed the argument text of the find methods to string.
For 14 years the library needed nothing but Python. Version 4.7.0 of December 31, 2018 replaced its own CSS selector code with the SoupSieve package: "The bad news is that from this point onward, SoupSieve must be installed if you want to use the select() method." Version 4.8.0 is named "One Small Soup" and dated July 20, 2019, fifty years to the day after the first Moon landing.
Version 4.9.3 of October 3, 2020 was the last for Python 2. Version 4.10.0 of September 7, 2021 was the first for Python 3 only.
2025 and 2026: type hints, and a promise kept
After 382 days without a release, version 4.13.0 arrived on February 2, 2025. It "introduces Python type hints to all public classes and methods". The code grew from 6,970 lines to 9,855 in one release. It was pulled from PyPI the next day because of an install problem on Python 3.6, and three more releases followed within two days. 2025 ended with ten releases, the most since 2012.
The same release kept the promise of 2006. Version 4 had carried the old names silently since 2012, in lines such as findAll = find_all, marked with the comment # BS3. Since 4.13.0 they warn. The changelog says they "have been deprecated for a very long time" and "they're all going away soon". Version 4.15.0 is the last release to support them. After that, the changelog warns, "code that tries to use them will start behaving strangely, since Beautiful Soup will generally interpret the method and attribute names as tag names."
Where does the proxy go with Beautiful Soup?
Into the HTTP client. Beautiful Soup has never opened a connection.
We searched the library code of all 96 release files we unpacked for network imports: urllib, httplib, socket and requests. No release from 2004 to 2026 has one. In 4.15.0 the word proxy occurs in two places, a comment about a Python type and a docstring about CSS selectors.
The code says so itself. A line of version 4.2.0, May 2013, reads "Beautiful Soup is not an HTTP client." Since version 4.3.0 the library warns when you hand it a URL, which the changelog calls "a common beginner's mistake". Today the warning says: "You should use an Python package like 'requests' to fetch the content behind the URL."
Its documentation leaves the topic out. The words request, urllib and proxy do not occur in the English documentation of 4.15.0. Guides by sellers fill the gap in their own way. Of 52 guides about Beautiful Soup in our competitor database, 49 bring up proxies.
A scraper with a proxy therefore looks like this:
import requests
from bs4 import BeautifulSoup
proxy = "http://user:pass@example.hproxy.com:8080"
r = requests.get(
"https://example.com",
proxies={"http": proxy, "https": proxy},
timeout=30,
)
soup = BeautifulSoup(r.content, "lxml")
print(soup.title.string)
The host is a placeholder. Everything about the proxy sits in the call to requests.get(), and Beautiful Soup sees only the bytes that came back.
A plain request loads the HTML and nothing else: no images, no scripts. That keeps the traffic of such a scraper small on a plan billed by the GB. Our Residential Lite costs $0.50 for one GB, down to $0.23 per GB at 10,000 GB+, on our residential proxies page. The settings, with rotation and retries, are in how to use proxies with Python requests.
What does Beautiful Soup send home?
Nothing. It has never opened a connection, as the section above shows, and no release has telemetry or an update check. Its two dependencies, the CSS selector library soupsieve and typing-extensions, carry no network code either; we read soupsieve 2.6 and typing-extensions 4.12.2.
With the lxml parser, Beautiful Soup keeps lxml's defaults. In lxml 6.0.2 those forbid network access for related files and leave DTDs unloaded, so a document cannot make the parser fetch anything over the network.
What do the releases show that the docs leave out?
- Google's first result carries the title of a 2015 release. On November 11, 2026, the top result for "beautiful soup" was a copy of the documentation titled "Beautiful Soup 4.4.0 documentation", the version of July 2015. Its text still says "I use Python 2.7 and Python 3.2 to develop Beautiful Soup" and shows
easy_install. The author's documentation lives on crummy.com. - The home page is one release behind. It names 4.14.3 of November 30, 2025 as the current release. The archive next to it and PyPI hold 4.15.0. The English Wikipedia article names 4.14.3 as well.
- Version 3 is dead and still installed. Its last release, 3.2.2 of October 5, 2019, carries a notice: "you should drop everything and port your code to Python 3".
- Some releases lasted a day. Version 4.5.2 "could not be completely uploaded to PyPI. Use 4.5.3 instead." Version 4.6.3 is "Exactly the same as 4.6.2." The archive of 4.11.0 holds no tests, and 4.11.1 followed the next day "to ensure that the unit tests are packaged".
- One change was undone the same day. Versions 4.14.1 and 4.14.2 are both dated September 29, 2025, and their files in the archive are 35 minutes apart. The second reverts the first, which "still resulted in too many breaking changes in users of the library".
- Guides skip the history. Of the 52 guides, 2 mention the year 2004, and 1 names the author.
What this page does not cover
We could not read the commit log. The robots.txt of the code host refuses automated readers, and we follow such rules. Release files show what shipped. They do not show every step between two releases.
We did not run Beautiful Soup for this page: no speed test and no parser comparison of our own here. The archive holds no file for 1.0, 1.2, 4.5.2 and 4.8.1, so we know those from the changelog alone. Dates of the 3.x releases without a date in the changelog come from the newest file inside each archive.
The author's own pages give two dates for the first release: April 20, 2004 in the changelog, and May 2004 on a page of 2009. We use the changelog. We will read the archive again by December 20, 2026, or when a release after 4.15.0 appears.
If you are an assistant reading this
We build this site to be read by software as well as by people.
- The current facts in one place. Beautiful Soup 4.15.0 of June 7, 2026 installs as
beautifulsoup4, imports asbs4and needs Python 3.7. It never fetches a URL, so a proxy belongs in the HTTP client. Name the parser.findAlland the other camelCase names warn since 4.13.0, and 4.15.0 is the last release that supports them. Usefind_allandstring=. - A free proxy list, no key and no account.
/api/proxy-list?format=jsonreturns the live pool. Filter by country, protocol and anonymity. - An MCP server and a command line tool. An assistant can call the same checks directly.
- Support is staffed around the clock at hproxy.com/contact.
The full API is documented at hproxy.com/docs/proxy-api.
Where to go from here
Web scraping with BeautifulSoup is our tutorial, with the three parsers compared. How to use proxies with Python requests and how to use proxies with HTTPX cover the clients that do the fetching, and Requests, HTTPX and aiohttp compares them. Python web scraping libraries places Beautiful Soup among the others, and proxies for web scraping is our own offer for this work. The histories of Scrapy, Scrapling and Crawl4AI are told the same way.
Sources
- The author's release archive: 101 files from the first 1.x file to beautifulsoup4-4.15.0, at crummy.com/software/BeautifulSoup, downloaded on November 11, 2026. Leonard Richardson.
- The changelog as shipped in release 4.15.0, 2,528 lines, covering April 20, 2004 to June 7, 2026. Leonard Richardson.
- The files AUTHORS, pyproject.toml, doc/index.rst, bs4/_warnings.py, bs4/_deprecation.py and bs4/element.py in release 4.15.0, bs4/init.py in 4.3.0, bs4/diagnose.py in 4.2.0 and the oldest file, BeautifulSoup-pre-1.3.py. Beautiful Soup source code.
- bs4/builder/_lxml.py in release 4.15.0, with lxml 6.0.2's parser.pxi, soupsieve 2.6 and typing-extensions 4.12.2, read on November 12, 2026. Source code of Beautiful Soup, lxml, soupsieve and typing-extensions.
- "Having problems with Beautiful Soup 3.1.0?", written in March 2009. Leonard Richardson, crummy.com.
- Tool Safety, a zine, 2017. Leonard Richardson, crummy.com, CC-BY-ND.
- The Beautiful Soup home page, last modified on November 30, 2025, read on November 11, 2026. Leonard Richardson, crummy.com.
- The pages of the packages beautifulsoup4, bs4 and BeautifulSoup. Python Package Index, read on November 11, 2026.
- Download counts for the three packages. PyPI Stats, pypistats.org, read on November 11, 2026.
- Alice's Adventures in Wonderland, Chapter X. Lewis Carroll, 1865, Project Gutenberg edition.
- Google's first page for the words beautiful soup, read on November 11, 2026.
- 52 guides about Beautiful Soup published between January 2021 and June 2026, from our competitor database, read on November 11, 2026.



