LangChain
LangChainFramework for building applications over language models, composing chains, agents, memory and a large ecosystem of integrations across model providers, vector stores and tools. It provides the primitives that LangGraph and the surrounding LangChain tooling build on, and is distributed as a set of Python packages.
LangSmith and the LangGraph Platform are sold separately as the managed layer; the framework packages are not feature-gated against them. Verified 2026-08-11 via the langchain-ai/langchain repository tree, the vendor product page and its pricing page.
Openness
5 high confidence- license
- MIT(langchain-core,langgraph,integrations)
- commercial
- LangSmith-platform/cloud
- source
- public
- core-gated
- ungated(every package in langchain-ai/langchain is MIT with no enterprise directory
MIT core with LangSmith sold separately. A full untruncated recursive tree of langchain-ai/langchain at master is 3,587 entries whose only license files are the root MIT and per-package copies under libs/, with no enterprise, ee or commercial path anywhere, and langchain.com/pricing sells only LangSmith - observability, evaluation, deployment - listing langchain, langgraph and deepagents separately under "Open Source Frameworks". That the vendor sells a paid product at all says nothing about the core, and a separate paid product beside a core with nothing withheld from it is not a gated core. This is scored on the langchain repo alone: langgraph is a separate product on this map, and it is gated by the Elastic-2.0 langgraph-api server.
- https://github.com/langchain-ai/langchain recorded 2026-08-15
Repository page for langchain-ai/langchain: 'MIT license' in the sidebar, 144.2k stars, public and unarchived. Establishes the licence and that the source is public.
- https://www.langchain.com/langchain recorded 2026-08-15
The vendor product page, which sells LangSmith and LangGraph Platform beside the open framework and claims no framework capability the MIT packages lack. That is the distinction between 5/open_source and 4/open_core, so it establishes core-gated.
- https://rvernica.github.io/2026/03/langchain-license recorded 2026-08-15
Third-party post examining LangChain's licensing across its packages. Corroborates the MIT reading; establishes no dimension on its own.
- https://api.github.com/repos/langchain-ai/langchain/git/trees/master?recursive=1 recorded 2026-08-11
Full untruncated recursive tree, 3,587 entries. Only license files are the root LICENSE and per-package copies under libs/: core, langchain, langchain_v1, model-profiles, seventeen partners/*, standard-tests, text-splitters. No enterprise/, ee/, commercial/ or proprietary/ path. Top-level directories are .devcontainer, .github, .vscode, libs.
- https://api.github.com/repos/langchain-ai/langchain recorded 2026-08-11
Repo metadata: default_branch master, language Python, license spdx_id MIT, size 588,758 KB, 143,996 stars.
- https://www.langchain.com/pricing recorded 2026-08-11
Page is titled 'LangSmith Plans and Pricing' and prices only LangSmith: Developer $0/seat, Plus $39/seat, Enterprise custom, plus metered LangChain Compute Units ($1.50/LCU) and Storage Units ($1.00/LSU) for Deployment, Fleet, Engine, Sandboxes and LLM Gateway. The nav lists langchain, langgraph and deepagents separately under 'Open Source Frameworks'. No paid tier of the langchain library itself.
Adoption
5 high confidence281,508,705 PyPI downloads in the trailing 30 days for the declared artifact, which bands at >10M, the top level; LangChain and LangGraph together have passed 1B cumulative downloads. PyPI is LangChain's primary channel and the package is the product rather than an SDK for a hosted service, so this measures the product itself.
- https://pypistats.org/api/packages/langchain/recent recorded 2026-08-13
281,508,705 downloads in the trailing 30 days for langchain
- https://www.langchain.com/langchain recorded 2026-08-13
LangChain presented as the open source framework, with LangSmith sold separately (corroborating, not the basis)
- https://grokipedia.com/page/LangChain recorded 2026-08-13
background on the framework and its ecosystem (corroborating, not the basis)
Capability
4 high confidenceBroadest LLM/agent orchestration ecosystem: chains, LangGraph stateful agent graphs, hundreds of model/tool integrations. No single benchmark for the framework; rated on feature breadth.
- https://github.com/langchain-ai/langchain recorded 2026-08-13
chains, agents and the integration ecosystem published in full under MIT; no benchmark figure
- https://www.langchain.com/langchain recorded 2026-08-13
framework positioning across chains, agents and integrations; no benchmark figure
Verified 2026-08-11