LangChain has raised $300 million at a $3 billion valuation. CapitalG, Alphabet’s growth fund, led the round again, with GIC and DFJ Growth coming in as new investors and IVP and Sapphire Ventures adding more. Salesforce Ventures, Insight Venture Partners, QuantumLight and MongoDB Ventures are also new, and Sequoia, Amplify and Benchmark put in more. Total funding now stands at $460 million.
That’s a big number for a company that started as an open source Python library for chaining LLM calls. The money isn’t really for the library anymore. It’s for LangSmith, the commercial platform LangChain has built around it, and for a bet on how enterprises will end up using AI agents.
The pitch: own your agents
CEO Harrison Chase’s argument is that agents will change how companies run, and that whoever controls those systems keeps the advantage. So companies should be free to swap models, keep their own data and infrastructure, and keep improving their agents, with governance tight enough to run them at scale. LangChain’s line for this is helping “every company own its intelligence.”
It’s a pointed message. The model labs would like enterprises to build on their own agent stacks. LangChain is selling the neutral layer in between, the one that lets you run Claude today, GPT tomorrow and a fine-tuned open model next quarter without rebuilding everything.
What LangSmith actually does
LangChain splits the platform into three parts.
Runtime. The open source frameworks (LangChain, LangGraph and Deep Agents) handle how agents are built and how context is managed. LangSmith Agent Server and Sandboxes run them, in the cloud or inside the customer’s own environment. Managed Agents and Fleet are the hosted option for teams that don’t want to operate any of it.
Improvement. Observability and Evaluation show what agents are doing and how well. LangSmith Engine is an agent inside the platform that hunts for errors, writes code and prompt fixes, and watches production traffic for regressions. Fine-Tuning turns agent traces into training data for smaller specialized models.
Governance. LLM and Tool Gateways sit between agents and everything they touch. They handle model budgets and fallbacks, protect sensitive data, and control which agents can use which tools and credentials.
Who’s using it
LangChain says more than 7,000 organizations use LangSmith. That includes startups like Harvey, Rippling and Listen Labs, and big names like Nvidia, Toyota and Salesforce. Clay monitors hundreds of millions of agent runs a month on it. Workday is the best stat in the announcement: it says the time to deploy a new agent dropped from five months to under a day.
CapitalG partner Mo Jomaa frames it as the move from LLM experiments to production agents, which he calls one of the biggest shifts in enterprise software right now.
What to watch
The governance and gateway pieces are where the money is. Observability tools are getting crowded, and every cloud and model vendor now ships its own agent tracing. Central control over model spend, data and tool access is a harder problem, and it’s the one CIOs and security teams actually lose sleep over. If LangSmith becomes the control plane for agents across vendors, $3 billion will look cheap. If the clouds bundle “good enough” versions for free, it gets harder.
LangChain plans to show new LangSmith features at a release event on November 20, 2026. Chase says it’s still early, and that building agents that work well is harder than it should be.