Browser Use Had 6,423,201 PyPI Downloads. Stagehand Had 8,582,033 on npm.

6,423,201. That is how many times the PyPI distribution browser-use was downloaded from October 10 through November 8, 2026, in the without_mirrors series.

stagehand had 1,909,034 downloads in the same window. Browser Use’s Python count is 3.36 times Stagehand’s, a gap of 4,514,167 file requests. The npm count points the other way. @browserbasehq/stagehand had 8,582,033 downloads over those dates. That is 1.34 times the PyPI total for browser-use, and 4.50 times the PyPI total for stagehand.

The November 9 install census does not include a browser-agent library. The November 5 star snapshot does not either. The product calendar is Browser and computer-use agents, the next 6 months. This note is the install row those pages leave out. Daily refresh of the broader framework set lives on the downloads tracker.

On PyPI, Browser Use leads. On npm, Stagehand leads.

The window is 30 calendar days, October 10 through November 8, inclusive. Each Python total is the sum of daily without_mirrors rows from pypistats. browser-use and stagehand were read at 13:40:30 GMT. skyvern and notte were read at 13:40:31 GMT. Stars are stargazers_count.

November 9 is in the table and outside the 30-day total. The week column is November 2 through November 8. In that week Browser Use was 3.80 times Stagehand on PyPI: 1,610,367 against 423,949.

A second 30-day window, October 11 through November 9, puts the jump inside the total. browser-use moves to 6,526,033. stagehand moves to 1,937,369. The ratio is 3.37. skyvern moves to 6,631. notte moves to 4,055. The PyPI rank is unchanged.

notte is in the table because its summary was read before the sums, and it is not in the rank of agent libraries. The summary calls it browser infrastructure. The PyPI license_expression is SSPL-1.0. nottelabs/notte had 2,059 stars at 13:42:08 GMT, and GitHub’s license field on that object was NOASSERTION. Treat 4,041 as the install count of an infrastructure package, not as a third browser agent that almost caught Skyvern.

PyPI summaries, read at 13:40:17 GMT: browser-use is “Browser agent on Pydantic AI: one persistent Python REPL on raw CDP, cloud or local Chrome.” stagehand is “Browserbase's SDK for building browser agents.” skyvern’s summary field was empty. Its description says it automates browser-based workflows with LLMs and computer vision, and it links to github.com/Skyvern-AI/skyvern. That repository’s license field was AGPL-3.0 at 13:39:45 GMT. The description says the same license. browser-use and stagehand were MIT on both PyPI and GitHub.

requires_python, same documents: browser-use is <4.0,>=3.11, stagehand is >=3.11, skyvern is <3.15,>=3.11.

The npm package named browser-use is not the library.

@browserbasehq/stagehand version 4.4.0 was published 2026-11-04T17:20:39.326Z. The PyPI file of the same version was uploaded 2026-11-04T17:06:39Z. Same version number, same morning, two registries. The repository field on the npm record is github.com/browserbase/stagehand.

@browserbasehq/sdk at 10,053,702 is the largest npm number in this pass, and it is the client for Browserbase’s hosted browsers. It is not Stagehand. Adding it to 8,582,033 would double-count a company and still would not count a person.

agent-browser at 8,357,760 is Vercel’s CLI, version 0.42.0 published 2026-11-10T00:40:05.816Z. A coding agent that shells out to a CLI is a different install from a library you import. It stays out of the Browser Use-versus-Stagehand rank for that reason. The number is printed so a weekly chart that leads with it is traceable.

The unscoped npm name browser-use had 8,969 downloads. 6,423,201 divided by 8,969 is 716. The official library’s Python users, measured as file requests, are not hiding on npm under that name.

Skyvern had no npm package under the name skyvern. That lookup returned 404 at 13:40:18 GMT.

Skyvern has the stars of a popular library and the downloads of a small one.

Skyvern-AI/skyvern had 23,378 stars at 13:39:45 GMT. stagehand had 25,992. The star gap is 2,614. The PyPI gap is not close: 1,909,034 against 6,386. Stagehand was downloaded 298.9 times as often.

Skyvern’s star count was 3.66 times its own 30-day downloads. Browser Use’s PyPI downloads were 53.29 times its star count. A sort by stars puts Skyvern next to Stagehand. A sort by PyPI file requests puts Skyvern next to notte.

The license is the other reason not to treat them as substitutes. Stagehand and Browser Use are MIT. Skyvern’s repository and its PyPI description are AGPL-3.0. An MIT service that imports an AGPL agent loop has a distribution question a star chart will not ask. This note does not give legal advice. It records the SPDX field and the description line.

Using the langchain total from the November 9 census, 155,918,257, the same dates and the same mirror filter, langchain was downloaded 24.27 times as often as browser-use. Browser agents are a smaller install base than the general agent SDK. Inside that smaller base, the registry decides the winner.

Browser Use 1.0 shipped after the window.

GitHub’s releases list, read at 13:39:52 GMT, starts with tag 1.0.1, published 2026-11-10T09:27:28Z, then 1.0.0, published 2026-11-10T02:04:48Z. PyPI’s latest field at 13:41:05 GMT was still 1.0.0, uploaded 2026-11-10T02:34:13Z. The releases map had 1.0.0 and 1.0.0rc5. It did not have 1.0.1.

pushed_at on browser-use/browser-use was 2026-11-10T09:29:07Z, two minutes after the 1.0.1 tag. None of that upload is inside a window that ends November 8. The 6,423,201 is the pre-1.0 package. A census repeated after November 10 will be the first one that can show whether 1.0 changed the daily rate. This one cannot.

November 9 was 304,218 for browser-use, up from 178,587 on November 8, and 56,303 for stagehand, up from 30,126. 1.0.0rc5 was uploaded at 23:42 UTC on November 9. A file that appears in the last minutes of a day does not explain the day’s total. The same jump, on a package that did not ship 1.0 that day, is stagehand. The headline ratio uses the window that stops on November 8.

What this means if you are choosing a browser-agent library

Import browser-use when the agent loop is Python and you want the distribution people already request. On this window that was 6,423,201 file requests, MIT, Python >=3.11 and <4.0, and the current release is the 1.0.0 that landed on PyPI the morning of November 10. Confirm 1.0.1 yourself before you pin it. It was on GitHub at 09:27 UTC and not on PyPI at 13:41 UTC.

Import @browserbasehq/stagehand when the loop is TypeScript. That package, not the PyPI name, is the larger Stagehand number: 8,582,033. The PyPI package is the same version, 4.4.0, and a smaller count. Public pages that say Stagehand “has more downloads” are reading this registry, or a mix of the two, and rounding. Public pages that say Browser Use leads are reading PyPI, or the 120,535 stars. Both sentences were true on November 10. They are not the same sentence.

Do not pick Skyvern because 23,378 stars sit near Stagehand’s 25,992. The install count was 6,386, and the license on the repository was AGPL-3.0. Do not pick the npm package browser-use and think you installed Browser Use. You installed an unofficial port with 8,969 downloads.