LangChain

LangChain began in late 2022 as an open-source side project by Harrison Chase, then a machine learning engineer at the startup Robust Intelligence [3]. Chase started it as a "nights-and-weekends side project, with no plans to build a company" [8], a single Python package of roughly 800 lines published from his personal GitHub account [16]. The repository was created on October 17, 2022 [6], and the first release followed on October 24, 2022 [14]. Chase and Gola built the first version of the framework together [4]. A month after the initial release, ChatGPT launched, and interest in the project grew sharply [16]; Chase and Gola "decided to follow the momentum and turn the project into a company" [4], starting it in February 2023 [16].

In April 2023 the company announced a $10 million seed round led by Benchmark [14][3][9], followed, per TechCrunch, a week later by a $25 million Series A led by Sequoia at a reported $200 million valuation [3][9]. (Sacra dates the Series A to February 2024 [2].) At the seed announcement, roughly six months after first release, the project had over 20,000 GitHub stars, 10,000 active Discord members, and more than 350 contributors [14].

The product line expanded over the following two years. LangSmith, a commercial observability platform, entered closed beta in July 2023, drawing 70,000 signups, and reached general availability by February 2024 [8][16]. LangGraph, an orchestration framework "inspired by the limitations of the initial langchain," began development in summer 2023 and launched in early 2024 [16]. On October 20, 2025, LangChain shipped 1.0 releases of both LangChain and LangGraph [1][16].

LangChain Inc. maintains a family of open-source frameworks plus a separate commercial platform, LangSmith [5][13][16].

The LangChain framework provides abstractions and integrations for models, tools, and agent loops, with a standard interface for models, embeddings, and vector stores [5][6]. It is MIT-licensed and distributed as Python and TypeScript packages [6][14]. The 1.0 release was rewritten to run on LangGraph's runtime, with a commitment of no breaking changes until 2.0; the 0.x line was preserved as langchain-classic with extended support [1][16].

LangGraph is a low-level orchestration framework and runtime for building, managing, and deploying long-running, stateful agents, offering durable execution, streaming, human-in-the-loop, and persistence [5]. It is inspired by Pregel and Apache Beam, and its public interface draws on NetworkX [5]. The company cites Klarna, Uber, and J.P. Morgan as users [5]. Deep Agents is a higher-level agent harness built on LangGraph for autonomous, long-horizon tasks [5][13]. Other components include the LangChain Expression Language, a declarative syntax that compiles to asynchronous code [2], and LangServe, which converts a LangChain application into a FastAPI server [2].

LangSmith is a closed-source platform for observability, evaluation, and monitoring of LLM applications [3], built separately from the open-source framework and neutral to both the underlying model and framework [16]. The company states it ingests over 1 billion events per day on LangSmith [13] and does not use customer data to train models [17]. Following an October 2025 relaunch, LangSmith encompasses Observability, Evaluation, Deployment (formerly LangGraph Platform), and a no-code Agent Builder in private preview [1]. Additions announced in 2026 include LangSmith Engine, an autonomous agent that clusters failures from production traces and proposes fixes; LangSmith Fleet, a no-code agent builder; Sandboxes, hardware-virtualized microVM execution environments; Managed Deep Agents; the LangSmith LLM Gateway, a governance layer enforcing spend limits and redacting sensitive data; and SmithDB, a Rust-built database backing core LangSmith workloads [15][5].

LangChain has raised $260 million in total [2]. In July 2025, TechCrunch reported the company was raising at an approximate $1 billion valuation led by IVP [3], and Forbes reported talks to raise $100 million at $1.1 billion [12]; Sacra records a completed $100 million Series B in July 2025 [2]. On October 20, 2025, LangChain announced a $125 million round at a $1.25 billion valuation led by IVP, which the company and Sacra label the Series B [1][2][9]. Existing investors Sequoia, Benchmark, and Amplify participated, alongside new investors CapitalG, Sapphire Ventures, ServiceNow Ventures, Workday Ventures, Cisco Investments, Datadog Ventures, Databricks Ventures, and Frontline [1][9].

LangChain operates a freemium B2B model: the open-source frameworks are free, and monetization relies on LangSmith, with usage-based pricing for API calls and seat-based pricing for collaboration features [2]. The Developer plan is free; the Plus plan costs $39 per seat per month; Enterprise pricing is custom, with self-hosted and hybrid deployment options aimed at regulated industries [17][2]. Four sources placed annual recurring revenue between $12 million and $16 million as of mid-2025 [3].

Named customers include Replit, Clay, Harvey, Rippling, Cloudflare, Workday, and Cisco [1], with LangGraph production deployments cited at LinkedIn, Uber, J.P. Morgan, and BlackRock [16]. The company states it works with 35% of the Fortune 500 [1][13]. LangChain and LangGraph together report 90 million monthly downloads [1], and the GitHub repository lists over 144,000 stars [6].

Competitors include vertically integrated model providers (OpenAI, Microsoft, AWS, and Google) that bundle agent frameworks into their platforms [2][3]; open-source orchestration frameworks such as LlamaIndex, Haystack, AutoGen, and CrewAI [2]; and AI development platforms such as Cursor and Windsurf [2]. TechCrunch reports LangSmith "currently leads the burgeoning LLM operations space," naming the smaller open-source Langfuse and Helicone as its competitors [3].

In summer 2023 the original package drew criticism over breaking changes, hidden prompts, package bloat, and dependency conflicts. Chase conceded the framework "traded power for ease of use," and the feedback motivated LangGraph and, ultimately, the 1.0 rewrite [16].

A Cloud Security Alliance research note published March 29, 2026 documents "a pattern of serious flaws documented throughout 2024 and 2025" across LangChain and LangGraph [7]. These include CVE-2025-68664 ("LangGrinch"), a critical serialization-injection vulnerability in langchain-core with a CVSS score of 9.3, and CVE-2025-64439, a remote code execution flaw in langgraph-checkpoint [7]. On March 27, 2026, Cyera Research published a coordinated disclosure named "LangDrained" identifying multiple high- and critical-severity vulnerabilities [7]. The CSA states patches are available for all documented vulnerabilities and urges they be treated as "urgent remediation actions" [7]. Microsoft's Defender Security Research Team has warned that frameworks such as LangChain "act as the operating system for AI agents," and that a single vulnerability in such a foundational layer "carries systemic risk" [19].

At its Interrupt conference in May 2026, LangChain announced LangSmith Engine in public beta, general availability of Sandboxes, Managed Deep Agents, the LLM Gateway in private beta, Deep Agents 0.6, and LangChain Labs, an applied research effort on continual learning for agents with partners Harvey, NVIDIA, Prime Intellect, Fireworks, and Baseten [15]. In June 2026 the company published its State of Agent Engineering report, based on 1,340 survey responses, finding 57.3% of respondents running agents in production, up from 51% the prior year [18].