LangChain Alternatives: Downloads vs. Newer AI SDKs
LangChain's JavaScript core grew 3.2× in 46 weeks, yet its share of framework installs fell from 32% to 19%. An early default can grow and still fade.
The signal
- npm@langchain/core had 5.7M downloads in the week of 24 Aug, 3.2× the week of 6 Oct 2025. The Vercel AI SDK had 23.6M.
- npmThe OpenAI Agents SDK grew from 0.11M to 1.6M weekly downloads, taking 5.2% of the three.
- PyPIIn Python, after the 25 Aug counting change, langchain moved +3.6% between the weeks of 31 Aug and 14 Sep; openai moved +7.8%.
LangChain's share of weekly npm framework downloads fell from 32% to 19% in 46 weeks, while its own downloads grew 3.2×.
Share of combined weekly npm downloads of @langchain/core, ai (Vercel AI SDK) and @openai/agents, week of 6 Oct 2025 vs week of 24 Aug 2026, in %.
We use @langchain/core for LangChain because every LangChain.js app installs it; the langchain package itself had 0.67 downloads per core download in the first week and 0.53 in the last, so adding it would double-count. Shares are of these three packages only. The OpenAI Agents SDK starts from a small base (0.11M a week), so its 15.0× growth says more about its youth than its size. Weeks with an npm gap day are dropped; the latest complete week ends 30 Aug. PyPI figures use only data after the 25 Aug 2026 counting change.
Show the data (3 rows)
| Package | Week of 6 Oct 2025 | Week of 24 Aug 2026 | Growth | Share then | Share now |
|---|---|---|---|---|---|
| @langchain/core | 1.8M | 5.7M | 3.2× | 32.1% | 18.5% |
| ai | 3.6M | 23.6M | 6.5× | 66% | 76.3% |
| @openai/agents | 0.11M | 1.6M | 15.0× | 1.9% | 5.2% |
The Mode read
LangChain's npm downloads are still rising. But JavaScript developers now have other places to start: Vercel's AI SDK, and OpenAI's own Agents SDK. Is LangChain keeping its place?
No. @langchain/core grew 3.2× in 46 weeks, to 5.7M a week. But the Vercel AI SDK grew 6.5× and the OpenAI Agents SDK 15.0× from a small base. LangChain's share of the three fell from 32% to 19%. Python shows the same drift on a short, post-break window: langchain +3.6% in two weeks, openai +7.8%.
This is how an early default loses its distribution. Existing projects keep installing it, so the total rises and the dashboard looks fine. New projects pick something else, and that only shows in share. For founders, the number that predicts the next year is share of new projects. For incumbents, a growing total is not proof of a moat.
Growing 3.2× while losing share is how a default fades.
What we would back: Developer tools that win the default for new projects in a fast-growing category, judged by share, not totals.