Harrison Chase is the co-founder and chief executive of LangChain, whose open-source frameworks and commercial tools help developers build dependable applications and agents around language models. His work has made context, orchestration, evaluation, and human oversight central concerns of AI engineering.
From machine learning to LangChain
Chase studied statistics and computer science at Harvard, led an entity-linking team at financial-technology company Kensho, and subsequently led a machine-learning team at Robust Intelligence. Those roles exposed him to problems that would shape his later work: grounding software in reliable information and testing whether models behave dependably outside controlled demonstrations.
In October 2022, he released LangChain as an open-source Python package. What began as a side project organizing emerging patterns for connecting language models to data and tools gained momentum after ChatGPT launched the following month. Chase and Ankush Gola founded LangChain as a company in February 2023.
The product expanded alongside the practical demands of deploying AI applications. LangSmith, introduced in beta in 2023, provides tracing, debugging, testing, and evaluation. LangGraph, launched in 2024, gives developers finer control over execution state, prescribed workflows, model-directed actions, and human intervention. Chase also acknowledged shortcomings in LangChain’s early abstractions, including hidden prompts, breaking changes, and unwieldy dependencies, and rebuilt the framework around a simpler agent loop. In October 2025, the redesigned LangChain 1.0 arrived alongside $125 million in financing at a $1.25 billion valuation.
- Context must be engineered. Instructions shape behavior, examples demonstrate desired outputs, retrieval supplies relevant facts, and fine-tuning changes learned behavior. Chase treats the ingestion, transformation, selection, and inspection of that information as essential to building context-aware applications.
- Workflows and agents belong on the same spectrum. Reliable systems can enforce mandatory steps while allowing models to choose tools, navigate uncertainty, and pursue longer-running tasks. LangGraph reflects that combination of deterministic structure and agent discretion.
- Evaluation and observability make agents governable. Chase advocates representative test datasets, complete execution traces, production feedback, and model-assisted assessment. Those tools help developers improve performance while giving enterprise reviewers a practical basis for understanding risks and approving deployment.
- Agent autonomy must justify its risks. His enterprise-agent framework weighs the value of successful work against failure probability and potential damage. Reversible changes, draft outputs, approval checkpoints, and human correction reduce those costs without requiring supervision of every step.
- Ambient agents need durable human control. Chase built an email-oriented agent to explore event-triggered background work, editable actions, clarifying questions, and recovery from earlier execution states. His writing on open agent harnesses and owned memory argues that users should retain accumulated context instead of becoming locked into closed systems. More recently, he has highlighted automating the agent-improvement loop, particularly evaluation and environment engineering.
