LangChain Skills
Official LangChain skills: choose between LangChain, LangGraph and Deep Agents, quickstarts, RAG, memory, human approval and eval engineering
- 1,302 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skills are instructions and references and run nothing on their own
- The agent code your assistant writes will need OpenAI or Anthropic keys in the environment
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/langchain-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skills are instructions and references and run nothing on their own
- The agent code your assistant writes will need OpenAI or Anthropic keys in the environment
This is third-party code. Review the repository files before installing.
What it does
A set of skills from LangChain Inc. for agents that write code on LangChain, LangGraph and Deep Agents. The starter skill ecosystem-primer helps pick a layer: Deep Agents for long tasks with planning, files, memory and subagents, LangGraph for custom control flow, LangChain for an agent with a fixed set of tools. Further skills cover each layer: create_agent and middleware, RAG, StateGraph and state reducers, checkpointers with thread_id, interrupts and human review. Separately there are six quickstarts for Python and TypeScript, langgraph-cli for the project lifecycle, eval-engineering for designing Harbor tasks and swarm for dispatching independent work in parallel.
Who it is for. For developers building agents on the LangChain stack who ask Claude Code, Cursor or Codex to write that code by current rules.
Good fit when
- You are starting an agent and do not know whether to pick LangChain, LangGraph or Deep Agents
- The agent needs human approval before an action and has to resume through Command
- You are building RAG: document loaders, embeddings and vector stores
- Graph state must survive a restart: checkpointers, thread_id, cross-thread memory
- You need Harbor eval tasks for your agent and a growing body of project knowledge
Not a fit when
- You need LangSmith traces, datasets and evaluators: LangChain keeps a separate langsmith-skills repository for those
- The agent is not built on LangChain, LangGraph or Deep Agents: the skills are tied to their APIs
- You need stable contents: the author warns the project is in early development and may change
- You want a ready MCP server with tools: this is instructions for the agent only
Example request
Use the ecosystem-primer skill and tell me what to build on for an agent that reads tickets, searches a knowledge base and waits for my approval before replying to a customerLimitations
The author marks the project as early development, so skill contents may change. Quickstarts default to the model anthropic:claude-sonnet-5 and ask for a provider key; the OpenAI and Anthropic APIs do not work from Russia without a VPN. eval-engineering requires Harbor, run locally in Docker or in a supported cloud environment. The install.sh script targets only Claude Code and Deep Agents CLI.
How to disable. Delete the skill folders from the project or home skills directory where the install put them. If you installed the Claude Code plugin, disable it from the /plugin menu.
FAQ
How do I install LangChain Skills in Claude Code?
As a plugin: /plugin marketplace add langchain-ai/langchain-skills, then /plugin install langchain-skills@langchain-skills. Alternatively: npx skills add langchain-ai/langchain-skills --agent claude-code --skill '*' --yes --global.
How is LangChain Skills different from langsmith-skills?
These skills cover building agents on LangChain, LangGraph and Deep Agents. The README points trace and dataset work in LangSmith to the separate langsmith-skills repository.
Which should I pick: LangChain, LangGraph or Deep Agents?
The ecosystem-primer skill fixes the order of checks. Planning, files over a long session, persistent memory or subagents point to Deep Agents. Custom control flow points to LangGraph, and an agent with a fixed set of tools uses create_agent in LangChain.
Can I install just one skill?
Yes, the --skill flag takes a name. The README installs only eval-engineering this way: npx skills add langchain-ai/langchain-skills --skill eval-engineering --yes.
Which keys do I need after installing?
The skills themselves ask for none. The agent you build with them needs the chosen provider's key: the README shows export OPENAI_API_KEY and export ANTHROPIC_API_KEY. Quickstarts ask for provider and model, defaulting to anthropic:claude-sonnet-5.
Discussion
No questions or reviews yet.
Related
SkillsA LangGraph 1.0 guide as a skill: ReAct agents, StateGraph, middleware, memory, MCP and RAG, so Claude Code writes LangChain code without outdated APIs
LangSmith skills
LangSmith Skills
Official LangChain skills for LangSmith: query and export traces, build eval datasets and write custom evaluators from your agent
A CLI skill manager for AI agents: install from repositories, translate between agent formats and deploy into their configs
A CLI installer and package manager for coding agent skills: one command installs skills into many runtimes and helps build your own
A CLI and MCP server that turns doc sites, repos, PDFs and videos into skills and RAG-ready data
A CLI and Python SDK to search, install, create and evaluate reusable agent skills from a shared library
README in short
The README opens with a warning about early development and a short description: skills for agents on LangChain, LangGraph and Deep Agents. Next come install routes (npx skills, the Claude Code plugin, the install.sh script with --deepagents and --global), API key setup and an eval-engineering section. It ends with the skill list grouped into getting started, quickstarts, Deep Agents, LangChain, LangGraph, evaluation and utilities.
README badge
Original SKILL.md
--- name: ecosystem-primer description: "INVOKE FIRST for any LangChain / LangGraph / Deep Agents agent building project before consulting other skills or writing any agent code. Required starting point for up to date info on framework selection (LangChain vs LangGraph vs Deep Agents vs hybrid composition), agent patterns, install, environment setup, and which skill to load next." --- <overview> LangChain Inc. maintains three layered open-source tools for building agents, plus LangSmith for observability. The stack, top-down: - **Deep Agents** (top layer, *harness*) — batteries-included toolkit built on LangChain + LangGraph. Ships with planning, file management, subagent spawning, and memory out of the box. - **LangGraph** (middle layer, *runtime*) — low-level orchestration for durable execution, custom control flow, and stateful workflows. LangChain agents run on top of LangGraph. - **LangChain** (bottom layer, *framework*) — abstractions for models, tools, and the agent loop. Provider-agnostic, easiest to start with. - **LangSmith** (cross-cutting) — observability and evaluation platform. Framework-agnostic; always recommended alongside any of the above. Higher layers depend on lower ones, but you don't need to use lower layers directly. Deep Agents gives you LangGraph's durable execution without writing graph code. LangChain gives you models and tools without managing graph edges. </overview> --- ## Step 1 — Choose Your Tool <decision-table> Evaluate these conditions in order and stop at the first match: 1. If the task needs planning, file management across a long session, persistent memory, subagent delegation, or on-demand skills → **Deep Agents** 2. Else, if the task needs custom control flow (deterministic loops, branching logic) → **LangGraph** 3. Else, if it's a single-purpose agent with a fixed set of tools → **LangChain** (`create_agent` function) 4. Else, if it's a pure model call, retrieval pipeline, or simple prompt chain with no agent loop → **LangChain** (direct model / chain) This is your **layer**. BUT you are not done: later in Step 4, you MUST load the layer-specific skill before writing any agent code. </decision-table> --- ## Tool Profiles <langchain-profile> ### LangChain — agent framework