Chains are the backbone of LangChain

They connect prompts, models, tools, memory, and logic to execute tasks step by step.
Instead of making a single LLM call, chains let you build multi-step reasoning, retrieval-augmented flows, and production-grade agent pipelines.

𝐇𝐞𝐫𝐞’𝐬 𝐚 𝐛𝐫𝐞𝐚𝐤𝐝𝐨𝐰𝐧 𝐨𝐟 𝐭𝐡𝐞 𝐦𝐨𝐬𝐭 𝐢𝐦𝐩𝐨𝐫𝐭𝐚𝐧𝐭 𝐭𝐲𝐩𝐞𝐬 𝐨𝐟 𝐜𝐡𝐚𝐢𝐧𝐬 𝐲𝐨𝐮 𝐧𝐞𝐞𝐝 𝐭𝐨 𝐤𝐧𝐨𝐰:

𝟏. 𝐋𝐋𝐌𝐂𝐡𝐚𝐢𝐧 (𝐁𝐚𝐬𝐢𝐜)
A straightforward chain that sends a prompt to the LLM and returns a result. Ideal for tasks like Q&A, summarization, and text generation.

𝟐. 𝐒𝐞𝐪𝐮𝐞𝐧𝐭𝐢𝐚𝐥 𝐂𝐡𝐚𝐢𝐧
Links multiple chains together. The output of one becomes the input of the next. Useful for workflows where processing needs to happen in stages.

𝟑. 𝐑𝐨𝐮𝐭𝐞𝐫 𝐂𝐡𝐚𝐢𝐧
Automatically decides which sub-chain to route the input to based on intent or conditions. Perfect for building intelligent branching workflows like routing between summarization and translation.

𝟒. 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦 𝐂𝐡𝐚𝐢𝐧
Allows you to insert custom Python logic between chains. Best for pre-processing, post-processing, and formatting tasks where raw data needs shaping before reaching the model.

𝟓. 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 𝐂𝐡𝐚𝐢𝐧𝐬
Combine retrievers with LLMs for grounded, fact-based answers. Essential for RAG systems where data retrieval must be accurate and context-aware.

𝟔. 𝐀𝐏𝐈 / 𝐒𝐐𝐋 𝐂𝐡𝐚𝐢𝐧
Connects external APIs or databases with LLM logic, enabling real-time queries or structured data processing before generating responses.

These chain types are what make LangChain powerful. They transform a single model call into dynamic, intelligent workflows that scale.