• Portfolio

Owning Your Intelligence

Our investment in LangChain.

November 13, 2026

by Caroline Fiegel and Tre Brown

• Founders: Harrison Chase (CEO), Ankush Gola (CTO)
• Sector: Developer Infrastructure / AI Agents
• Location: San Francisco, CA

The Opportunity

Three years ago, building with large language models meant prototyping chatbots. Today, enterprises are putting agents to work on real operations: claims processing, customer support, internal knowledge, and software development. Those agents will be sold as products and used to run internal operations alike. The hard part is no longer getting a model to answer a question. It’s making an agent reliable: knowing what it saw, what it did, why it did it, and whether the last change made it better or worse.

Model capability continues to improve, commoditizing general intelligence, but in order for a company to use AI to run critical parts of its business or embed it in the product it sells, generic intelligence won’t be enough. Companies need to own their intelligence: the model choice, the harness that turns model output into action, the context and memory that make a generic model specific to a business, and the feedback loop that makes the 100th interaction better than the first.

That framing mirrors what we hear from enterprise buyers. The teams moving agents from pilot to production need model and cloud neutrality, cost and access controls, and observability that withstands an audit. Few vendors cover that full stack.

The Solution

LangChain has been at the center of this shift since the beginning. The open-source langchain package shipped in October 2022, a month before ChatGPT, and became the default way to build LLM applications. As the industry’s needs matured, so did the product line: LangGraph for controllable, durable agent runtimes, Deep Agents as a harness for long-horizon tasks, and LangSmith, the commercial platform that brings agent runtime, improvement, and governance together. Agent Runtime (Agent Server, Managed Agents, and Sandboxes) runs agents securely at scale, in the cloud or in the customer’s own environment. Agent Improvement (Observability, Evaluation, LangSmith Engine, and LangSmith Fine-Tuning) tests, monitors, and improves them. Agent Governance (LLM and Tool Gateways) gives teams central control over how agents use models, tools, and data. 

The newest pieces close the loop. LangSmith Engine clusters production problems into prioritized issues, finds the root cause in traces and agent code, and proposes a fix for review. LangSmith Managed Agents, launched in August, bundles the harness with the runtime, sandboxes, context management, evals, memory, and auth so developers can deploy a production agent with one command.

Critically, LangSmith is framework-neutral and model-neutral. It works whether a team builds on LangChain’s open-source frameworks or on another stack entirely, which is what enables LangChain to serve as a platform, not just a framework.

Why We’re Backing LangChain

We’ve had the privilege of knowing LangChain CEO Harrison Chase since early 2023, and have been continually impressed by his ability to absorb the frontier and make it usable by the everyday developer. That shows in an open-source community that has become the de facto starting point for each research wave: retrieval-augmented generation (RAG), tool use, stateful agents, and harness design. That community has translated into real enterprise traction. More than 7,000 organizations rely on LangSmith, from the fastest-growing startups to Fortune 500 companies.

We’ve seen firsthand in our enterprise community how value on LangChain’s platform compounds. Every agent run on LangSmith produces traces, and those traces become the raw material for everything that follows: evals that catch regressions, Engine fixes grounded in real failures, and, increasingly, models tuned on a team’s own production data. That accumulated learning belongs to the customer and travels with them when the model underneath changes, which is the strength of a neutral platform. The longer a company runs on LangSmith, the more it has invested there and the harder the platform is to replace. We believe this is the layer where advantage accumulates, and LangChain is the company best positioned to own it. 

What’s Ahead

LangChain’s mission is to enable every company to own its intelligence. As LangChain continues to extend the platform toward governed runtimes, tuned evaluators, and bring-your-own-cloud deployments for regulated buyers, it will become increasingly critical to the enterprise stack.

We’re proud to join CapitalG, GIC, IVP, Sapphire Ventures, DFJ Growth, Sequoia, Benchmark, and the rest of the syndicate in backing Harrison, Ankush, and the LangChain team. 

Welcome to Salesforce Ventures, LangChain!