LangChain

LangChain builds the agent engineering platform and open-source frameworks developers need to ship reliable agents faster. Founded in early 2023 by Harrison Chase and Ankush Gola, the company has grown from a single Python package into a full-stack platform covering observability, evaluation, deployment, and no-code agent creation. According to the company's About page, LangChain works with 35% of the Fortune 500, has crossed 1 billion open-source downloads, and ingests over 1 billion events per day on LangSmith.

What It Is

LangChain is an agent engineering company whose commercial product, LangSmith, covers the full agent development lifecycle: tracing and observability, evaluation, deployment infrastructure, and a no-code agent builder called Fleet. Alongside LangSmith, LangChain maintains three open-source frameworks — LangChain (quick-start agents with any model provider), LangGraph (low-level control for production agents), and Deep Agents (long-running, autonomous agents for complex tasks). LangSmith is framework-agnostic and integrates with any agent stack via Python, TypeScript, Go, and Java SDKs, as well as native OpenTelemetry support.

The LangSmith Platform in Depth

LangSmith is organized into five major capability areas:

• Observability — Structured tracing breaks each agent run into a timeline of steps, with native support for popular frameworks, OpenTelemetry, multi-turn message threading, and AI-driven analytics to surface patterns across traces.
• Evaluation — Capture production traces, convert them into test cases, and score agents with LLM-as-judge evaluators, multi-turn evals, human annotation queues, and online/offline scoring. Eval calibration with human feedback closes the loop between production signals and iterative improvement.
• Deployment — A purpose-built agent server provides memory, conversational threads, durable checkpointing, horizontal scaling, type-safe streaming, and native A2A and MCP protocol support. Supports human-in-the-loop interactions, background agents, and cron scheduling.
• Fleet — A no-code agent builder that lets non-developers describe tasks in plain language and turn them into recurring agents that act across daily tools. Fleet supports bring-your-own models, first-party integrations, MCP server extensions, and exports agent files for pro-code development.
• Sandboxes — Ephemeral, isolated compute environments for safely running agent-generated code, with configurable TTLs, snapshot-and-fork for parallel branches, port tunneling, and a Sandbox CLI.

Interrupt 2026: New Product Launches

At its Interrupt 2026 conference, LangChain announced a wave of new products that extend the LangSmith platform:

• LangSmith Engine — An autonomous issue-detection layer that clusters production failures into prioritized issues, traces root causes in code, and proposes fixes for developer review. Described by LangChain as a way to "surface and diagnose undetected issues autonomously to improve agents faster."
• SmithDB — A purpose-built database layer for agent state and memory, designed to support the persistence requirements of long-running agents.
• Managed Deep Agents — A hosted version of the Deep Agents framework, enabling teams to run highly autonomous, long-horizon agents without managing their own infrastructure.
• LangSmith Sandboxes — Ephemeral, isolated sandboxes for safely executing agent-generated code at scale, available on the Plus plan and above.
• LLM Gateway — A unified gateway layer for routing, managing, and governing LLM calls across providers within the LangSmith platform.
• LangSmith Context Hub — A centralized store for managing context, prompts, and retrieval assets used by agents, enabling teams to version and share context across projects.

Open-Source Frameworks

LangChain's open-source work underpins its commercial platform and keeps the company close to emerging agent-building patterns:

• LangChain — The original framework for quickly wiring together LLM calls, tools, and chains with any model provider.
• LangGraph — A lower-level framework for building production agents that require determinism, branching logic, and fine-grained state control.
• Deep Agents — A newer harness for building highly autonomous agents capable of planning, using subagents, and leveraging file systems for complex, long-running tasks.

The company publishes the open-source projects to stay ahead of best patterns, then incorporates learnings into LangSmith.

Deployment Model and Enterprise Options

LangSmith is available as a cloud SaaS (US or EU regions), a hybrid deployment (SaaS control plane with a self-hosted data plane), or fully self-hosted within a customer's own VPC. Enterprise customers get custom SSO, RBAC, support SLAs, team trainings, architectural guidance, and access to deployed LangChain engineers. The company also offers a Startup Plan with discounted rates and credits for VC-backed early-stage companies. LangChain is headquartered in San Francisco with offices in New York, Boston, and Amsterdam, and is backed by IVP, Benchmark, and Sequoia.