LangChain raises $300M at a $3B valuation to turn LangSmith into the operating system for agents

The agent era has a plumbing problem. Everyone can demo an agent that books a flight or drafts a report; far fewer can run one in production for a year without it quietly degrading, overspending, or doing something it shouldn’t. LangChain, the startup whose open-source frameworks became the default scaffolding for building agents, just raised $300 million to sell the fix — and investors priced it at $3 billion.

The Series C, announced on November 13, was led by CapitalG, Alphabet’s growth fund and an existing LangChain investor. New investors GIC and DFJ Growth took significant stakes, alongside existing backers IVP and Sapphire Ventures. Salesforce Ventures, Insight Venture Partners, QuantumLight, and MongoDB Ventures also joined, while Sequoia, Amplify, and Benchmark increased their positions. The round brings LangChain’s total funding to $460 million.

The round, in numbers#

The valuation step is the part worth pausing on. Roughly a year ago, LangChain’s Series B priced the company at $1.25 billion. The new $3 billion figure is about 2.4 times that — a steep climb for a company that sells developer tooling, and a sign that investors now treat agent infrastructure as a category of its own rather than a feature of the model labs. The Wall Street Journal first reported the round on November 12; the company confirmed it the next day.

What LangSmith actually does#

LangSmith is the commercial half of LangChain. The open-source LangChain and LangGraph frameworks are what developers use to build agents; LangSmith is what companies use to run them without losing control. It is deliberately framework-agnostic — it works with agents built on any stack, not just LangChain’s own — and it is organized around three jobs: running agents reliably, improving them, and governing them.

Agent Runtime covers deployment: LangSmith Agent Server and Sandboxes, plus managed services for teams that want LangChain to operate the whole thing. Agent Improvement covers observability and evaluation, plus LangSmith Engine, an in-platform agent that hunts for errors, writes fixes, and watches production traffic for regressions, and fine-tuning from agent traces. Agent Governance covers the controls — LLM and tool gateways that manage model budgets and fallbacks, protect sensitive data, and govern which tools an agent may touch.

The customer list is doing a lot of the persuading. More than 7,000 organizations use LangSmith, according to the company, from startups like Harvey, Rippling, and Listen Labs to Nvidia, Toyota, and Salesforce. Clay, the sales-automation company, uses it to monitor hundreds of millions of agent runs a month. Workday says LangSmith cut the time to deploy a new agent from five months to under a day. None of those figures are independently audited — they are the company’s numbers — but the names on the list are real enterprises, and that is the point of the raise: LangChain is no longer selling to hobbyists.

Why investors are paying up for agent plumbing#

There is a pattern in this funding cycle. The model labs raise tens of billions; the companies that make the models usable in production raise hundreds of millions at multi-billion valuations. Nominal raised $240 million at $3 billion days earlier to bring AI to hardware engineering; Fractile raised $720 million at $7 billion for inference chips. LangChain’s round sits in the same current: the value is migrating from the model itself to the layer that makes it reliable, observable, and governable.

That is also the bet behind LangChain’s pitch — “own your intelligence.” The argument is that companies should be free to swap models, control their data and infrastructure, and keep improving their agents, rather than renting intelligence from a single provider’s black box. It is a pitch aimed squarely at enterprises that have watched model prices, capabilities, and terms change under them all year. Whether “owning” your agent stack is a real moat or a slogan will be decided by how sticky LangSmith becomes once it is embedded in a customer’s deployment pipeline.

What to watch#

First, the November 20 event. LangChain says it will share several new LangSmith capabilities there — the first concrete test of what this $300 million actually buys. Second, the numbers the company did not disclose: revenue, profitability, and how the money will be split between product and go-to-market. A $3 billion valuation for a developer-tools company invites questions about the revenue multiple, and LangChain did not volunteer an answer.

Third, the competitive field. Agent platforms are crowded — from the model labs’ own tooling to a wave of startups selling observability, evals, and governance. LangChain’s advantage is distribution: its open-source frameworks put it in the room before the sales conversation starts. The risk is the reverse: that the frameworks stay free and beloved while the paid platform faces cheaper, narrower rivals. The next year will show which side of that ledger the $300 million lands on.

Sources: LangChain — LangChain Raises $300M Series C at $3B Valuation (Nov. 13, 2026); Business Wire — LangChain Raises $300M to Enable Every Company to Own Their Intelligence (Nov. 13, 2026); Unite.AI — LangChain Secures $300M Series C at $3B Valuation for Agent Governance (Nov. 13, 2026). Prior coverage: DXC × LangChain: governed AI agents for mission-critical IT, with 95% faster incident resolution.