LangChain & LangGraph
Your agent’s research. Ready to present.
Add Encelade tools to your LangChain or LangGraph agent through MCP. Generate a deck from its research, notes, and data.
Quickstart
Load it as tools
MultiServerMCPClient pulls the Encelade tools over streamable HTTP and hands them to any LangGraph agent. Pass your API key in the headers and you are done.
- Works with LangGraph prebuilt and custom agents
- No hand-written tool schemas
- Streamable HTTP transport with Bearer auth
- Combine with your existing LangChain tools
# pip install langchain-mcp-adapters langgraph "langchain[anthropic]"
import asyncio
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
async def main():
client = MultiServerMCPClient(
{
"encelade": {
"url": "https://www.encelade.ai/api/mcp",
"transport": "streamable_http",
"headers": {"Authorization": "Bearer YOUR_API_KEY"},
}
}
)
tools = await client.get_tools()
agent = create_react_agent("anthropic:claude-sonnet-4-5", tools)
result = await agent.ainvoke(
{"messages": "Turn these Q1 sales notes into a client-ready deck: ..."}
)
print(result)
asyncio.run(main())How it works
From source material to shareable link.
- 01
Install the adapter
Add langchain-mcp-adapters to your environment. It bridges any MCP server into LangChain-native tools.
- 02
Register Encelade
Point MultiServerMCPClient at /api/mcp with your Bearer token and call get_tools() to load the presentation tools.
- 03
Generate from the agent
Your agent now calls generate_project as part of its reasoning and returns a shareable presentation URL in its answer.
Let the agent poll
Generation runs asynchronously: generate_project returns a run ID, and the agent calls get_generation_run until it is done. A ReAct agent handles this naturally across turns — just make sure your run isn't capped to a single tool call.
Questions
LangChain
& Encelade.
How do I add Encelade as a LangChain tool?
Use langchain-mcp-adapters. Register Encelade's /api/mcp endpoint in MultiServerMCPClient with your Bearer token, call get_tools(), and pass the result to any LangGraph agent.
Does it work with LangGraph?
Yes. The loaded tools work with create_react_agent and with custom LangGraph graphs — Encelade tools behave like any other LangChain tool node.
Do I need to write tool schemas?
No. The MCP adapter generates the tool definitions from the server, so you don't hand-write input schemas or HTTP calls.