Give your coding agent guidance for building LangChain agents, LangGraph workflows, and Deep Agents.
LangChain Skills packages development guidance for coding agents working with LangChain, LangGraph, and Deep Agents. It covers choosing a framework, building local examples, adding retrieval and persistent state, and evaluating agent behavior. In Claude Code, relevant skills can load automatically, and you can select one from the slash menu. The guides direct the agent to start new agent projects with ecosystem-primer. Running the generated examples requires credentials for your chosen model provider. The project is in early development, so its APIs and skill content may change.
Describe the tools, control flow, memory, and planning your application needs. The ecosystem primer guides the coding agent through choosing LangChain, LangGraph, or Deep Agents and selecting the next development skill.
Ask for a Python or TypeScript quickstart with your preferred model provider. The quickstart skills direct the agent to build an example in a new directory, configure the provider key, and run it.
When an agent needs answers grounded in documents, the RAG skill guides work on loading files, splitting text, creating embeddings, and retrieving content from a vector store.
For a LangGraph workflow that must wait for a reviewer, the human-in-the-loop and persistence skills cover interrupts, saved state, and resuming the same conversation.
Use the Deep Agents guides when your application needs task planning, file access, persistent memory, or delegation to specialized subagents.
The eval-engineering skill guides the coding agent through inspecting a repository and optional traces, proposing a task for human review, and building and auditing a runnable Harbor evaluation. Harbor runs need Docker locally or a supported cloud environment.
Showing all 22
Choose an agent framework, prepare the environment, and select the next skill.
Choose packages and manage Python or TypeScript dependencies across the LangChain ecosystem.
Build and run a local Python weather agent from the official quickstart.
Build a local TypeScript weather agent; requires Node.js 22 or later.
Build and run a local Python calculator agent with the Graph API.
Build and run a local TypeScript calculator agent with the Graph API.
Build a local Python research agent using the model provider’s web search.
Build a TypeScript research agent with provider web search; requires Node.js 22+.
Build LangChain agents with tools, middleware, and structured output.
Add tool-call approvals, custom middleware, and structured output to LangChain agents.
Build document retrieval with text splitting, embeddings, and vector stores.
Define LangGraph state, nodes, edges, branching, and streaming execution.
Route LangGraph workflows with typed decision models and calibrated thresholds.
Pause LangGraph execution for human review and resume with a decision.
Save workflow state, retain conversation history, and configure cross-thread memory.
Configure a Deep Agent’s planning, file management, tools, and skill loading.
Configure temporary or persistent file storage and route between storage backends.
Coordinate specialized subagents, task plans, and human approvals in Deep Agents.
Process independent items in parallel; requires the QuickJS interpreter’s swarm_task tool.
Scaffold, develop, build, and deploy LangGraph applications with the CLI.
Build and deploy managed agents in LangSmith; public beta on US Cloud only.
Design reviewed evaluation tasks, audit Harbor runs, and preserve reusable project knowledge.
No matching results.
The Claude Code plugin route adds the marketplace and installs its plugin. The Skills CLI routes install the collection’s skills for the current project or across your projects; they do not install plugin hooks, commands, or MCP servers.
Inside a Claude Code session on your computer, add the LangChain Skills marketplace.
/plugin marketplace add langchain-ai/langchain-skillsOpen the plugin’s installation panel, choose its scope, and confirm installation.
/plugin install langchain-skills@langchain-skillsRead the installation summary. If the plugin is active now, continue. If a reload is needed, close the panel to apply it. If Claude Code leaves the change pending after a prompt-cache warning, restart Claude Code to load the plugin.
Type a forward slash in the prompt and look for skills under the langchain-skills prefix, such as ecosystem-primer.
Install scope
In a terminal, install the collection’s skills for Claude Code across your projects.
npx skills add langchain-ai/langchain-skills --agent claude-code --skill '*' --yes --globalIn a terminal at your project root, install the collection’s skills for Claude Code in this project.
npx skills add langchain-ai/langchain-skills --agent claude-code --skill '*' --yesStart a new Claude Code session in that project, or any project for a global install. Type a forward slash in the prompt and look for ecosystem-primer and the other installed skills.
Install scope
In a terminal, install the collection’s skills for GitHub Copilot in VS Code across your projects.
npx skills add langchain-ai/langchain-skills --agent github-copilot --skill '*' --yes --globalIn a terminal at your project root, install the collection’s skills for GitHub Copilot in VS Code in this project.
npx skills add langchain-ai/langchain-skills --agent github-copilot --skill '*' --yesFor a project install, open that project in VS Code. In VS Code, select Open Customizations from the Chat view and inspect the available skills for Copilot. Look for ecosystem-primer. You can also type a forward slash in the chat input to find available skills.
Install scope
In a terminal, install the collection’s skills for GitHub Copilot across your projects.
npx skills add langchain-ai/langchain-skills --agent github-copilot --skill '*' --yes --globalIn a terminal at your project root, install the collection’s skills for GitHub Copilot in this project.
npx skills add langchain-ai/langchain-skills --agent github-copilot --skill '*' --yesStart a new Copilot CLI session in that project, or any project for a global install, and ask it to list the available skills. Check that ecosystem-primer appears.
Install scope
In a terminal, install the collection’s skills for Codex across your projects.
npx skills add langchain-ai/langchain-skills --agent codex --skill '*' --yes --globalIn a terminal at your project root, install the collection’s skills for Codex in this project.
npx skills add langchain-ai/langchain-skills --agent codex --skill '*' --yesFor a project install, open that project in Codex. In Codex CLI or the IDE extension, type a dollar sign in the prompt and look for ecosystem-primer in the skill picker. In the desktop app, open Skills in the sidebar. If the new skills do not appear, restart Codex and check again.