LangChain Community Plugin for Claude Code

Unofficial community plugin - not affiliated with LangChain, Inc.

Build AI applications with LangChain and LangGraph. This plugin provides comprehensive guidance for agents, RAG pipelines, and LCEL chains with best practices baked in.

Features

Skill	Use Case	Triggers
langgraph	AI agents with tools, state, HITL	"build an agent", "add tools to agent"
langchain-rag	RAG pipelines with vector stores	"build RAG pipeline", "set up vector store"
langchain-chains	LCEL chains for summarization, extraction	"create summarization chain", "extract data"

Installation

Via Marketplace (Recommended)

# Add the marketplace
/plugin marketplace add github:Codeblockz/langchain-community-plugin

# Install the plugin
/plugin install langchain-community

Direct Installation

# Install from GitHub (supports updates)
claude plugins add https://github.com/Codeblockz/langchain-community-plugin

To update the plugin later:

claude plugins update langchain-community

Alternatively, install from a local path:

claude plugins add /path/to/langchain-community-plugin

Local Development

To test the plugin during development without installing:

claude --plugin-dir /path/to/langchain-community-plugin

This loads your plugin directly. Restart Claude Code to pick up changes as you develop.

Verifying the plugin loaded:

• Run /help to see commands listed under langchain-community:
• Run /agents to see the reviewer agents
• Ask Claude to "build an agent" to trigger the langgraph skill

Debugging:

• Use claude --debug to see plugin loading errors
• Check YAML frontmatter syntax in skill/command files

Skills

langgraph (Agents)

Build AI agents that can use tools, maintain state, and interact with humans.

When to use:

• Autonomous agents that decide which tools to call
• Multi-step reasoning with tool use
• Human-in-the-loop approval workflows
• Stateful conversations with memory

Example triggers:

• "Build an agent that can search the web and write files"
• "Create a ReAct agent with custom tools"
• "Add human approval before tool execution"

langchain-rag (RAG Pipelines)

Build retrieval-augmented generation pipelines with any vector store.

Supported vector stores:

• InMemoryVectorStore (prototyping)
• FAISS (local, file-based)
• Chroma (local with server option)
• pgvector (PostgreSQL)
• Pinecone (managed cloud)
• Qdrant (open source, cloud)
• Weaviate (open source, cloud)

When to use:

• Question answering over documents
• Semantic search applications
• Knowledge bases with retrieval
• Document-grounded chatbots

Example triggers:

• "Build a RAG pipeline to answer questions about my docs"
• "Set up FAISS vector store with persistence"
• "Load PDFs and create searchable index"

langchain-chains (LCEL Chains)

Build data processing pipelines using LangChain Expression Language.

When to use:

• Summarization (single doc or multiple)
• Data extraction with structured output
• Text classification
• Translation
• Any prompt → LLM → parser workflow

Example triggers:

• "Create a summarization chain for long documents"
• "Extract entities from text into Pydantic models"
• "Build a classification pipeline with confidence scores"

Commands

Agents

Command	Description
/langchain-community:new-agent	Scaffold a new agent (create_agent or StateGraph)
/langchain-community:add-tool	Add a tool to an existing agent

RAG

Command	Description
/langchain-community:new-rag	Scaffold a RAG pipeline with your choice of vector store

Chains

Command	Description
/langchain-community:new-chain	Scaffold a chain (summarization, extraction, classification, translation)

Agents (Code Reviewers)

langgraph-reviewer

Automatically reviews LangGraph agent code for common mistakes:

• Missing checkpointers for persistence
• Wrong state annotation types
• Missing reducers for list fields
• Recursion limit issues
• Incorrect interrupt placement

Triggers proactively after writing agent code, or invoke explicitly: "Review my agent code"

rag-reviewer

Automatically reviews RAG pipeline code for common mistakes:

• Embedding dimension mismatches
• Missing metadata preservation
• Unhandled empty results
• Suboptimal chunk sizes
• Missing FAISS deserialization flag

Triggers proactively after writing RAG code, or invoke explicitly: "Review my RAG pipeline"

Quick Start Examples

Create an Agent

You: Build an agent that can search Wikipedia and save notes

Claude: [Uses langgraph skill, scaffolds agent with tools]

Create a RAG Pipeline

You: I need to build a Q&A system over my company docs using FAISS

Claude: [Uses langchain-rag skill, creates pipeline with FAISS persistence]

Create a Chain

You: Create a chain to extract contact info from emails

Claude: [Uses langchain-chains skill, creates extraction chain with Pydantic]

Requirements

• Python 3.10+ (3.11 recommended)
• Core: pip install "langchain>=1.0" "langchain-openai>=1.0"
• Agents: pip install langgraph
• RAG: Vector store specific (see skill docs)
  • FAISS: pip install faiss-cpu
  • Chroma: pip install langchain-chroma
  • pgvector: pip install langchain-postgres psycopg[binary]
  • Pinecone: pip install langchain-pinecone pinecone-client

Plugin Structure

langchain-community-plugin/
├── .claude-plugin/
│   └── plugin.json
├── skills/
│   ├── langgraph/           # Agent patterns
│   ├── langchain-rag/       # RAG pipelines
│   └── langchain-chains/    # LCEL chains
├── commands/
│   ├── new-agent.md
│   ├── add-tool.md
│   ├── new-rag.md
│   └── new-chain.md
├── agents/
│   ├── langgraph-reviewer.md
│   └── rag-reviewer.md
└── LICENSE

Version

v0.1.0 - Added RAG and Chains skills, renamed to langchain-community