With its LangChain Series C, the company behind the agent frameworks of the same name has raised 300 million US dollars at a valuation of 3 billion US dollars. The round, announced on 13 November 2026, is led by existing investor CapitalG. The money is meant mainly for LangSmith, the commercial platform for running, improving and governing AI agents. For developers, the valuation matters less than the direction of a tool that serves as the foundation of many agent projects.
LangChain Series C: The key facts of the round
In the announcement by CEO Harrison Chase, LangChain names CapitalG as lead investor. GIC and DFJ Growth join as new investors with significant participation, and existing investors such as IVP and Sapphire Ventures take part again. The company also welcomes Salesforce Ventures, Insight Venture Partners, QuantumLight and MongoDB Ventures as new backers. Sequoia, Amplify and Benchmark increased their commitment in this round.
For context on the valuation: according to TechCrunch, it stood at 1.25 billion US dollars about a year ago, when LangChain raised 125 million US dollars in a round led by IVP. The new valuation is therefore more than twice as high. LangChain does not disclose revenue or profit figures in the announcement.
From open-source project to agent platform
LangChain began in late 2022 as a side project by Harrison Chase. In early 2023 he founded the company together with Ankush Gola. The early libraries handled tasks that language models of the time struggled with, such as web search, API calls and database access. Today, the major model providers deliver many of these capabilities themselves, so LangChain has shifted its focus to running and managing agents.
The lead investor gives concrete usage figures in its post on the investment. According to CapitalG, the open-source projects have passed one billion downloads in total. At the time of the Series B, the two core packages LangChain and LangGraph had around 90 million monthly downloads combined. Today the projects see more than 350 million downloads a month, now including Deep Agents, the open-source harness for long-running autonomous work. These figures come from the investor and have not been independently verified.
CapitalG describes the way of working that LangChain calls “agent engineering” as a discipline of its own. Because agents are non-deterministic, getting one into production takes a loop of testing, measuring and refining, with product, engineering, data science and AI teams involved. The investor compares this agent lifecycle of building, testing, deploying, monitoring, iterating and governing with the way the software development lifecycle took shape a generation ago, when source control, CI/CD and monitoring became standard.
What LangSmith covers today
LangChain describes LangSmith as a framework-agnostic platform built on three pillars:
- Agent Runtime: the Deep Agents, LangGraph and LangChain frameworks are meant to offer flexibility across models and control over context. LangSmith Agent Server and Sandboxes run agents in the cloud or in the customer’s own environment. Managed Agents and Fleet are fully managed services.
- Agent Improvement: observability and evaluation show how agents behave and measure quality. LangSmith Engine is an in-platform agent that finds errors, writes code or prompt fixes and monitors production traffic for regressions. Fine-Tuning turns agent traces into training data for specialised models.
- Agent Governance: LLM and tool gateways centrally manage model budgets and fallbacks, protect sensitive data and control which tools agents may use, with which authentication.
According to CapitalG, LangSmith runs in LangChain’s cloud, self-hosted, or in the customer’s own AWS, Azure or Google Cloud account. This autumn the team shipped Engine v2, Fine-Tuning and bring-your-own-cloud for Azure and Google Cloud, among other things.
Customers and use cases, according to the company
More than 7,000 organisations use LangSmith, LangChain says. It names start-ups such as Harvey, Rippling and Listen Labs and large enterprises such as Nvidia, Toyota and Salesforce. Two examples are described in more detail: Clay says it monitors hundreds of millions of agent runs per month with LangSmith. Workday uses it to run and govern agents across the company and has cut the time to deploy a new agent from five months to under a day. Both claims come from the announcement and cannot be verified from the outside.
Analysis: neutrality as a selling point
LangChain’s central message is that every company should “own its intelligence”. In practice, that means companies should choose models freely, control their data and infrastructure, and keep improving their agents. CapitalG also justifies the investment by pointing out that model labs and hyperscalers ship their own agent tooling, while large enterprises want a neutral layer across several providers. The investor adds that the top spot on the leaderboards changed hands several times this year and that the best open-weight models trail the frontier by only a few months.
Whether that neutrality holds up remains to be seen. TechCrunch notes that plenty of younger vendors make similar promises. At the same time, model providers keep expanding their own agent platforms and payment features, as recently seen with ChatGPT Wallet and Sierra’s open protocol Poppy 1.0. For teams working with LangGraph or Deep Agents today, the funding changes nothing about their code in the short term. It does, however, increase the pressure to grow the commercial product.
What developers should take away
For teams that run agents in production, or plan to, a few practical points follow:
- The frameworks remain the entry point; new capabilities land mainly in LangSmith. Anyone using only the open-source libraries should watch the changelogs to see which capabilities stay open.
- Evaluations are the core of switching models. Teams that want to swap models need test sets that make regressions measurable — regardless of which platform vendor they use.
- Governance features such as gateways for models and tools are becoming standard. Teams building them in-house should weigh the effort against platform offerings.
- For self-managed data, the operating model still matters: LangChain’s cloud, self-hosting or your own cloud account have different consequences for data protection and cost.
On Friday, 20 November, LangChain plans to present further LangSmith capabilities at a release event. How agents can be evaluated systematically is covered in our piece on Jev AI as an LLM-as-a-judge alternative and the LangChain study.




