/langchain — LangChain SDK Skill

/langchain — LangChain SDK Skill

You now have full access to the LangChain SDK via MCP tools in the langchain-bridge server.

Available MCP tools

Tool	What it does
mcp__langchain-bridge__langchain_chain	Run a LangChain LCEL chain (prompt → LLM → output). Single call, fast.
mcp__langchain-bridge__tavily_search	LangChain TavilySearch — raw web results as JSON

How to use

When the user invokes /langchain, read their request and:

1. Simple Q&A / text tasks → call langchain_chain(question=..., system_prompt=...)
2. Web search needed → call tavily_search(query=...)
3. Chained tasks → call tools in sequence, passing outputs forward

LCEL chain system_prompt tips (keep short — token budget)

• Summarise: "Summarise concisely in bullet points."
• Classify: "Classify the input into one of these categories: ..."
• Extract: "Extract structured data as JSON."
• Translate: "Translate to Spanish."

Example invocations

User: /langchain summarise this text: [text] → langchain_chain(question="Summarise: [text]", system_prompt="Summarise concisely in bullet points.")

User: /langchain search latest AI news → tavily_search(query="latest AI news 2025")