Build

The LangChain open source stack provides the building blocks you need to design, test, and ship agents. Each layer is yours to configure: the model, the tools, the context and memory an agent works from, and the harness around the model loop.

Choose your starting point

Deep Agents, LangChain, and LangGraph share the same open source stack, so choose based on how much control you need. For managed hosting or no-code, start with Managed Deep Agents or Fleet:

Python
TypeScript

Deep Agents

Build agents for complex, long-running tasks. A complete agent harness with planning, subagents, a virtual filesystem, and long-term memory built in. The fastest way to start.

Managed Deep Agents

Build deep agents with managed deployment infrastructure. You write the agent; LangSmith runs the harness and runtime.

LangChain

A minimal, configurable agent framework. Compose exactly what you need from models, tools, prompts, and middleware.

LangGraph

Low-level orchestration for stateful, long-running agents: durable execution, streaming, memory, and human-in-the-loop.

Prefer no-code?

Build and run agents without writing code using LangSmith Fleet.

Use a ready-made agent

Deep Agents Code

Open source terminal coding agent (dcode) built on the Deep Agents SDK. Switch models mid-session, customize skills and memory, and approve shell execution from the CLI.

Explore

Python
TypeScript

Integrations

Connect to model providers, vector stores, retrievers, and other components.

Learn

Follow tutorials and conceptual guides for common agent patterns and use cases.

Reference

API references, error codes, release notes, and migration guides.

Contribute

Contribute documentation, code, and integrations to the LangChain ecosystem.