ai-framework

LangChain pricing & features

The framework for building LLM apps — chains, agents, RAG, LangGraph

Dominant OSS framework for LLM apps. LangChain (chains), LangGraph (stateful agents), LangSmith (obs). Python + JS/TS. MIT-licensed core.

Data sourced from vendor documentation · Last updated May 2026

LangChain website ↗Docs ↗

Pricing

Limits

Features

• Agents — Tool-using agents with reasoning loops. · docs
• Chains (LCEL) — LangChain Expression Language — pipe primitives into chains. · docs
• Checkpointers (LangGraph) — Persist agent state to SQL, Mongo, Redis, Postgres.
• Document Loaders — 150+ loaders for PDF, HTML, Notion, Google Drive, S3, GitHub, etc.
• Human-in-the-loop — Pause agent for approval, then resume.
• LangGraph — Stateful graph-based agent runtime. Durable, replayable, human-in-the-loop. · docs
• LangGraph Platform — Managed hosting for LangGraph agents with state persistence.
• LangGraph Studio — Desktop IDE for debugging agent graphs.
• LangServe — Deploy chains as FastAPI endpoints.
• Memory — Buffer, summary, entity, vector memory stores.
• Output Parsers — Structured JSON, Pydantic schemas, function calling.
• Prompt Templates — Templating + partial filling + output parsers.
• RAG (Retrieval-Augmented Generation) — Standard patterns + 50+ retrievers. · docs
• Streaming — First-class streaming at every layer.
• Subgraphs — Compose agent graphs hierarchically.
• Text Splitters — Recursive, token, semantic splitters for chunking.
• Tools — 400+ pre-built tools (web search, code, databases, APIs).
• Vector Store Integrations — 60+ vector DBs (Pinecone, Chroma, Weaviate, PGVector, Qdrant, Milvus).

Developer interfaces

Related ai-framework platforms

Compare LangChain with