langchain-dependencies

Which packages to install, and at which versions, to start or repair a LangChain project in Python or TypeScript — including what to pin down and what has since been replaced.

WHEN TO USE

SCOPE

USE FOR

Pick the exact packages and minimum versions needed for a new LangGraph or Deep Agents project.
Decide whether to pin, use latest, or avoid a package like langchain-community that skips semver.
Fix a broken install caused by a missing peer dependency or an outdated community package.
Find the current dedicated integration package to replace a deprecated langchain-community import.

NOT FOR

Do not start a new project on legacy LangChain 0.3 — it only gets security patches until December 2026.

This is a saved copy of the skill's SKILL.md instructions. It may differ from the upstream file or the version you install.

In this document (7)

The LangChain ecosystem is split into focused, independently-versioned packages. Understanding which packages you need — and their version constraints — prevents incompatibilities and keeps upgrades predictable.

Key principles:

LangChain 1.0 is the current LTS release. Always start new projects on 1.0+. LangChain 0.3 is legacy maintenance-only — do not use it for new work.
langchain-core is the shared foundation: always install it explicitly alongside any other package.
langchain-community (Python only) does NOT follow semantic versioning; pin it conservatively.
LangGraph vs Deep Agents: choose one orchestration approach based on your use case — they are alternatives, not a required stack (see Framework Choice below).
Provider integrations (model, vector store, tools) are installed separately so you only pull in what you use.

Environment Requirements

Requirement	Python	TypeScript / Node
Runtime minimum	Python 3.10+	Node.js 20+
LangChain	1.0+ (LTS)	1.0+ (LTS)
LangSmith SDK	>= 0.3.0	>= 0.3.0

Framework Choice

Pick one agent orchestration layer. You do not need both.

Framework	When to use	Core extra package
LangGraph	Need fine-grained graph control, custom workflows, loops, or branching	`langgraph` / `@langchain/langgraph`
Deep Agents	Want batteries-included planning, memory, file context, and skills out of the box	`deepagents` (depends on LangGraph; installs it as a transitive dep)

Both sit on top of langchain + langchain-core + langsmith.

Core Packages

Python — always required

Package	Role	Min version
`langchain`	Agents, chains, retrieval	1.0
`langchain-core`	Base types & interfaces (peer dep)	1.0
`langsmith`	Tracing, evaluation, datasets	0.3.0

Python — orchestration (pick one)

Package	Use when	Min version
`langgraph`	Building custom graphs directly	1.0
`deepagents`	Using the Deep Agents framework	latest

Python — model providers (pick the one(s) you use)

Package	Provider
`langchain-openai`	OpenAI (GPT-4o, o3, …)
`langchain-anthropic`	Anthropic (Claude)
`langchain-google-genai`	Google (Gemini)
`langchain-mistralai`	Mistral
`langchain-groq`	Groq (fast inference)
`langchain-cohere`	Cohere
`langchain-fireworks`	Fireworks AI
`langchain-together`	Together AI
`langchain-huggingface`	Hugging Face Hub
`langchain-ollama`	Ollama (local models)
`langchain-aws`	AWS Bedrock
`langchain-azure-ai`	Azure AI Foundry

Python — common tool & retrieval packages

These packages have tighter compatibility requirements — use the latest available version unless you have a specific reason not to.

Package	Adds	Notes
`langchain-tavily`	Tavily web search (`TavilySearch`)	Dedicated integration package; prefer latest
`langchain-text-splitters`	Text chunking utilities	Semver, keep current
`langchain-community`	1000+ integrations (fallback)	NOT semver — pin to minor series
`faiss-cpu`	FAISS vector store (local)	Via `langchain-community`; use latest
`langchain-chroma`	Chroma vector store	Dedicated integration package; prefer latest
`langchain-pinecone`	Pinecone vector store	Dedicated integration package; prefer latest
`langchain-qdrant`	Qdrant vector store	Dedicated integration package; prefer latest
`langchain-weaviate`	Weaviate vector store	Dedicated integration package; prefer latest
`langsmith[pytest]`	pytest plugin for LangSmith	Requires langsmith >= 0.3.4

langchain-community stability note: This package is NOT on semantic versioning. Minor releases can contain breaking changes. Prefer dedicated integration packages (e.g. langchain-chroma, langchain-tavily) when they exist — they are independently versioned and more stable.

TypeScript — always required

Package	Role	Min version
`@langchain/core`	Base types & interfaces (peer dep)	1.0
`langchain`	Agents, chains, retrieval	1.0
`langsmith`	Tracing, evaluation, datasets	0.3.0

TypeScript — orchestration (pick one)

Package	Use when	Min version
`@langchain/langgraph`	Building custom graphs directly	1.0
`deepagents`	Using the Deep Agents framework	latest

TypeScript — model providers (pick the one(s) you use)

Package	Provider
`@langchain/openai`	OpenAI (GPT-4o, o3, …)
`@langchain/anthropic`	Anthropic (Claude)
`@langchain/google-genai`	Google (Gemini)
`@langchain/mistralai`	Mistral
`@langchain/groq`	Groq (fast inference)
`@langchain/cohere`	Cohere
`@langchain/aws`	AWS Bedrock
`@langchain/azure-openai`	Azure OpenAI
`@langchain/ollama`	Ollama (local models)

TypeScript — common tool & retrieval packages

Package	Adds	Notes
`@langchain/tavily`	Tavily web search (`TavilySearch`)	Dedicated integration package; prefer latest
`@langchain/community`	Broad set of community integrations	Use sparingly; prefer dedicated packages
`@langchain/pinecone`	Pinecone vector store	Dedicated integration package; prefer latest
`@langchain/qdrant`	Qdrant vector store	Dedicated integration package; prefer latest
`@langchain/weaviate`	Weaviate vector store	Dedicated integration package; prefer latest

@langchain/core must be installed explicitly in yarn workspaces and monorepos — it is a peer dependency and will not always be hoisted automatically.

Minimal Project Templates

Minimal dependency set for a LangGraph project (provider-agnostic).

# requirements.txt
langchain>=1.0,<2.0
langchain-core>=1.0,<2.0
langgraph>=1.0,<2.0
langsmith>=0.3.0

# Add your model provider, e.g.:
# langchain-openai
# langchain-anthropic
# langchain-google-genai

Minimal package.json dependencies for a LangGraph project (provider-agnostic).

{
  "dependencies": {
    "@langchain/core": "^1.0.0",
    "langchain": "^1.0.0",
    "@langchain/langgraph": "^1.0.0",
    "langsmith": "^0.3.0"
  }
}

Minimal dependency set for a Deep Agents project (provider-agnostic).

# requirements.txt
deepagents            # bundles langgraph internally
langchain>=1.0,<2.0
langchain-core>=1.0,<2.0
langsmith>=0.3.0

# Add your model provider, e.g.:
# langchain-anthropic
# langchain-openai

Minimal package.json dependencies for a Deep Agents project (provider-agnostic).

{
  "dependencies": {
    "deepagents": "latest",
    "@langchain/core": "^1.0.0",
    "langchain": "^1.0.0",
    "langsmith": "^0.3.0"
  }
}

Adding Tavily search and a vector store to a LangGraph project.

# requirements.txt
langchain>=1.0,<2.0
langchain-core>=1.0,<2.0
langgraph>=1.0,<2.0
langsmith>=0.3.0

# Web search
langchain-tavily          # use latest; partner package, semver

# Vector store — pick one:
langchain-chroma          # use latest; partner package, semver
# langchain-pinecone      # use latest; partner package, semver
# langchain-qdrant        # use latest; partner package, semver

# Text processing
langchain-text-splitters  # use latest; semver

# Your model provider:
# langchain-openai / langchain-anthropic / etc.

Adding Tavily search and a vector store to a LangGraph project.

{
  "dependencies": {
    "@langchain/core": "^1.0.0",
    "langchain": "^1.0.0",
    "@langchain/langgraph": "^1.0.0",
    "langsmith": "^0.3.0",
    "@langchain/tavily": "latest",
    "@langchain/pinecone": "latest"
  }
}

Versioning Policy & Upgrade Strategy

Package group	Versioning	Safe upgrade strategy
`langchain`, `langchain-core`	Strict semver (1.0 LTS)	Allow minor: `>=1.0,<2.0`
`langgraph` / `@langchain/langgraph`	Strict semver (v1 LTS)	Allow minor: `>=1.0,<2.0`
`langsmith`	Strict semver	Allow minor: `>=0.3.0`
Dedicated integration packages (e.g. `langchain-tavily`, `langchain-chroma`)	Independently versioned	Allow minor updates; use latest
`langchain-community`	NOT semver	Pin exact minor: `>=0.4.0,<0.5.0`
`deepagents`	Follow project releases	Pin to tested version in production

Breaking changes only happen in major versions (1.x → 2.x) for all semver-compliant packages. Deprecated features remain functional across the entire 1.x series with warnings.

Prefer dedicated integration packages over langchain-community. When a dedicated package exists (e.g. langchain-chroma instead of langchain-community's Chroma integration), use it — dedicated packages are independently versioned and better tested.

Community tool packages (Tavily, vector stores, etc.) should be kept at latest unless your project requires a locked environment. These packages frequently release compatibility fixes alongside LangChain/LangGraph updates.

Environment Variables

All keys are read from the environment at runtime. Set only the keys for services you actually use.

# LangSmith (always recommended for observability)
LANGSMITH_API_KEY=<your-key>
LANGSMITH_PROJECT=<project-name>   # optional, defaults to "default"

# Model provider — set the one(s) you use
OPENAI_API_KEY=<your-key>
ANTHROPIC_API_KEY=<your-key>
GOOGLE_API_KEY=<your-key>
MISTRAL_API_KEY=<your-key>
GROQ_API_KEY=<your-key>
COHERE_API_KEY=<your-key>
FIREWORKS_API_KEY=<your-key>
TOGETHER_API_KEY=<your-key>
HUGGINGFACEHUB_API_TOKEN=<your-key>

# Common tool/retrieval services
TAVILY_API_KEY=<your-key>          # for Tavily search
PINECONE_API_KEY=<your-key>        # for Pinecone

Common Mistakes

Never start a new project on LangChain 0.3. It is maintenance-only until December 2026.

# WRONG: legacy, no new features, security patches only
langchain>=0.3,<0.4

# CORRECT: LangChain 1.0 LTS
langchain>=1.0,<2.0

langchain-community can break on minor version bumps — it does not follow semver.

# WRONG: allows minor-version updates that may be breaking
langchain-community>=0.4

# CORRECT: pin to exact minor series
langchain-community>=0.4.0,<0.5.0

Also consider switching to the equivalent dedicated integration package if one exists (e.g. langchain-chroma instead of the community Chroma integration).

Community tool packages like langchain-tavily and vector store integrations release compatibility fixes alongside LangChain updates. Using an old pinned version can cause import errors or broken tool schemas.

# RISKY: old pin may be incompatible with LangChain 1.0
langchain-tavily==0.0.1

# BETTER: allow latest within the current major
langchain-tavily>=0.1

Many tools that used to live in langchain-community now have dedicated packages with updated import paths. Always prefer the dedicated package import.

# WRONG — deprecated community import path
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_community.tools import WikipediaQueryRun
from langchain_community.vectorstores import Chroma
from langchain_community.vectorstores import Pinecone

# CORRECT — use dedicated package imports
from langchain_tavily import TavilySearch                  # pip: langchain-tavily (TavilySearchResults is deprecated)
from langchain_community.tools import WikipediaQueryRun  # no dedicated pkg yet
from langchain_chroma import Chroma                       # pip: langchain-chroma
from langchain_pinecone import PineconeVectorStore        # pip: langchain-pinecone

To find the current canonical import for any integration, search the integrations directory: https://python.langchain.com/docs/integrations/tools/

Each entry shows the correct package and import path. If a dedicated package exists, use it — the community path may still work but is considered legacy.

@langchain/core is a peer dependency — it must be in your package.json, especially in monorepos.

// WRONG: missing @langchain/core (breaks in yarn workspaces / strict hoisting)
{
  "dependencies": {
    "@langchain/langgraph": "^1.0.0"
  }
}

// CORRECT: always list @langchain/core explicitly
{
  "dependencies": {
    "@langchain/core": "^1.0.0",
    "@langchain/langgraph": "^1.0.0"
  }
}

Python 3.9 and below are not supported by LangChain 1.0.

# Verify before installing
import sys
assert sys.version_info >= (3, 10), "Python 3.10+ required for LangChain 1.0"

Node.js below 20 is not officially supported.

# Verify before installing
node --version   # must be v20.x or higher

Account or credentials required

2 FOR YOU

Items without a tag are required.

01
Get an API key for your model provider
Obtain and set the API key environment variable for whichever model provider you use, e.g. ANTHROPIC_API_KEY, OPENAI_API_KEY, or GOOGLE_API_KEY.
You set: a key for each model provider you use, such as ANTHROPIC_API_KEY, OPENAI_API_KEY or GOOGLE_API_KEY
02
Optional Get a LangSmith API key for tracing
Set LANGSMITH_API_KEY (and optionally LANGSMITH_PROJECT) to enable LangSmith tracing and evaluation, recommended for any project.
You set: LANGSMITH_API_KEY
You set: LANGSMITH_PROJECT

Listed as free: Connected services may require an account or charge fees.

This skill ships inside the LangChain Agent-Building Skills plugin.

Install just this skill

Run this outside your AI tool, using Node.js 22.20.0 or newer. Skills CLI 1.7.2 skips skill selection; prompts cancel installation without an interactive terminal. If choosing Project when asked about installation scope, open your project's folder before running this.

Install checklist

Check selected tools if a picker appears; Claude Code may be unchecked under Additional agents. ↑/↓ moves, Space toggles selections, Enter confirms.
Without a tool picker, check Installing to: for your intended tool. Stop if it names a different tool.
If asked, choose Symlink (Recommended) for shared files or Copy to all agents for separate copies; review the summary before confirming installation.
Install the find-skills skill? defaults to Yes, adding another skill for all projects. Choose No if you don't want that extra installation.

Skills CLI 1.7.2 sends install data to skills.sh by default, including the source, skill names, target tools and file details; GitHub sources are reported only when GitHub confirms the repository is public, while other remote sources can also be reported. Confirmed-public GitHub sources also trigger a security-audit request. Before installing, set the environment variable DO_NOT_TRACK=1 or DISABLE_TELEMETRY=1 to disable both.

npx skills add langchain-ai/langchain-skills --skill "langchain-dependencies" --full-depth

After installing for Claude Code, open it in the installation project if you chose Project; Global makes the skill available across projects on this computer. Once Claude Code loads the skill, its description helps Claude decide when to use it.

If installation created a top-level skills folder that did not exist when your Claude Code session started, run /reload-skills in that session. Run it again after later changes in that folder.

Install the whole plugin

This route installs the LangChain Agent-Building Skills plugin package selected for your AI tool.

Claude Code

Add LangChain's marketplace, then install the plugin

Type both lines inside a Claude Code session, not at your shell prompt. The first adds LangChain's marketplace, which is this repository; the second opens the details of the langchain-skills plugin, and Claude Code installs it once you pick one of the install options shown. Afterwards, LangChain's README asks you to set OPENAI_API_KEY or ANTHROPIC_API_KEY, for whichever of those model providers you build with.

/plugin marketplace add langchain-ai/langchain-skills

/plugin install langchain-skills@langchain-skills

Open plugin page ↗

What's inside

Files included with this skill.

View on GitHub