Early AI darling LangChain is now a unicorn with $125 million in new funding

The startup, which has created an open-source framework for connecting AI applications to real-time data, hopes its tools will become the virtual building blocks that companies use to unleash a plethora of AI agents — while its investors believe the company has the potential to become as successful as other foundational digital infrastructure companies like Crowdstrike (for cybersecurity) and Datadog (to monitor data).

The round, which was reported to have been completed over the summer, was led by IVP, with participation from existing investors Sequoia and Benchmark and new backers including CapitalG and Sapphire Ventures, Service now Ventures, Workday Ventures, Cisco Ventures, Datadog, Databricks, and Frontline. LangChain says its tools are already being used by AI teams at companies like Cisco, Replit, and Clay. CloudflareThe day of work and service is now.

The company says building trustworthy AI agents—systems that can think, act, and use tools on users’ behalf—is still very difficult. “Today, it is easy to create prototypes for dealers but difficult to ship,” Lang Chen wrote in a press release announcing the round. “Any agent input or change can lead to a host of unknown outcomes.”

The solution, the company says, is a new approach that blends product, engineering, and data science, which it calls agent engineering. The company positions itself as the connective tissue of the agent era — not just holding connectors together, but providing the full lifecycle of tools developers need to build, deploy, and monitor agents in production. For example, a company like ServiceNow might use LangChain to connect an LLM to its internal knowledge base and use it to run workflows or track performance.

LangChain was started in late 2022 as an open source project by Harrison Chase, then an engineer at Robust Intelligence, just weeks after its launch. OpenAI has released ChatGPT. She pioneered the idea of ​​“strings” – the building blocks that connect large language models to external tools and data sources in a sequence, allowing them to take actions rather than just generate text. A simple string might allow an AI to receive a user’s question, call a web search API, summarize the results, and return an answer, steps grouped together like links. It was an instant hit: “It was so crazy,” Chase recalls. “I didn’t know I was leaving my previous job. I had no idea what I was going to do next.”

The project that became the startup LangChain, which Chase co-founded with Ankush Gula, turned out to be a developer darling. That’s because it solved one of the most pressing problems in the early days of large language models: models couldn’t access real-time information or perform actions like searching the web, calling APIs, or interacting with databases. The LangChain framework allows developers to build these capabilities into their LLM applications, and adoption has skyrocketed. The San Francisco startup raised a $10 million seed round led by Benchmark in April 2023, announced a $25 million Series A in 2024 led by Sequoia, valuing the company at $200 million.

However, since then, the market has become crowded with other companies offering similar tools, such as LlamaIndex and Haystack, while OpenAI, Anthropic, Google It now provides built-in capabilities that were previously differentiators for LangChain.

To stay ahead of the curve, LangChain has expanded its product lineup, including LangSmith, a monitoring, monitoring, evaluation, and deployment platform specifically designed for LLM applications and agents. Since its launch last year, LangSmith has grown in popularity, with LangChain keeping some of its early products open source while creating proprietary platforms.

Langchain did not provide details about its financials, its spokesperson believes TechCrunch reported in July Which pegged its annual recurring revenue at between $12 million and $16 million was “low compared to where we are today.” Although the company is not profitable, Langechen is “fairly efficient in spending” compared to high-growth, venture capital-backed startups, the spokesperson said.

IVP Tom Lovero, who led the investment, said the company had “huge conviction” in Chase and the company’s potential from the beginning. “Two years ago, the question was whether an open source project like Lang Chen could become a major business,” he said. “We’ve seen Harrison and Ankush boldly take the first important steps on that journey,” including building multiple products that customers want.

Lovero said he sees LangChain as potentially being as successful as companies like Crowdstrike and Datadog, which have become indispensable for taming the complexity of cybersecurity and cloud infrastructure, respectively. LangChain is betting that it can become the layer that makes AI agents reliable and observable enough for companies to trust, turning today’s messy prototypes into business-critical systems. “It seems increasingly certain that agents are very important for the future,” he said. “And if you think so, then agent architecture is going to be very important.”

Chase acknowledges that the proxy platform scene is already crowded, but argues that LangChain’s breadth and neutrality will give it staying power. “There are a lot of players,” he said. “I would say we have 500 competitors and zero competitors at the same time.” He expects most companies will eventually use multiple agent platforms, many of which, like ServiceNow, will be powered under the hood by LangChain.

IVP’s Loverro confirmed that Langchain is already enjoying strong revenue and adoption and large companies like Cisco and Workday are relying on LangChain. He says there will be competition, “but it’s yet to be determined if it matters.” If investors are right, LangChain could become the indispensable layer that powers the age of proxies — just as CrowdStrike and Datadog did for the last generation of infrastructure.