LangChain raises $300 million, sets November 20 for LangSmith feature reveal

LangChain raised $300 million in a Series C that values the AI agent software company at $3 billion, it announced on November 13. CapitalG, Alphabet’s growth investment arm, led the round. LangChain also set November 20 as the date it will show new capabilities for LangSmith, the platform teams use to run and evaluate agents.

For developers already running agents, LangSmith provides a record of the models and tools called during a task, so teams can investigate a bad response and test a fix against the same example. Companies can choose where to run their agents and how to handle those records. LangChain’s financing announcement gives existing users a date for the next feature briefing. The company has yet to name those features or give dates for customer access.

CapitalG first invested in October 2025, when IVP led a $125 million Series B at a $1.25 billion valuation. LangChain’s latest valuation is 2.4 times that figure. New investors GIC and DFJ Growth joined the Series C alongside existing backers including IVP and Sapphire Ventures. The company put its total funding at $460 million.

Where agents run and where teams review them

The open-source framework that gave LangChain its name helps developers connect models to data and tools. LangSmith deals with what happens after those connections are in use. A trace can capture the model requests, retrieved material and results of tool calls in a single run. If a user reports a bad answer, an engineer can use that record to see where the agent picked up the wrong information.

The hosting choice determines where some of those records can go. LangChain’s documentation says Agent Servers deployed in its cloud send traces to its hosted LangSmith service. In a hybrid setup, the customer supplies the infrastructure that runs the agent and can turn off LangSmith tracing or send traces to the hosted service. A self-hosted deployment can send traces to a LangSmith instance the customer runs. The pricing page lists hybrid and self-hosted deployments as enterprise options.

LangChain says more than 7,000 organizations use LangSmith. Among them is Workday, which runs agents on its own infrastructure and uses LangSmith to trace and evaluate them, according to a November 10 case study. Workday’s platform team put common authentication and access controls in place so product teams could develop an agent’s behavior on top of them. The company used to spend four to five months building a prototype. With the shared platform, a team could get one running in staging in under a day. That result measures how quickly a prototype reached a test environment.

Turning failures into tests for the next release

LangSmith lets a team put examples from live traffic into a test dataset. Running the same requests against two versions can show whether a new model or tool has broken behavior customers rely on. In production, where a reference answer often isn’t available, teams can use evaluators to flag responses that fall outside their quality criteria. Those cases can inform the next round of pre-release tests.

LangSmith Engine takes on part of that review work. It groups recurring failures from traces and can suggest a test or a correction. With access to the team’s source repository, it can propose a change to code or a prompt on GitHub or GitLab. An October update added automated tests for the changes Engine proposes. A team member can review the change and test results before deciding whether to merge it. Engine can also suggest adding the failed run to a dataset that will be used to test future versions.

LangSmith’s single-seat Developer plan currently includes 5,000 base traces a month. Plus lists 10,000 monthly base traces at $39 per seat per month, with additional charges depending on usage. The pricing page lists Engine on Plus and Enterprise, with metered Engine use on Plus.

The November 20 event starts at 9 a.m. Pacific time. In the event listing, LangChain says co-founder and CEO Harrison Chase will discuss how companies measure the impact of their agents and manage cost and performance when agents run continuously.