I am building a data quality pipeline using LangGraph and Langflow. I want to connect Langflow agents to Snowflake via its native MCP server.

Langflow native MCP Tools component only accepts static headers (key/value form fields). It does not expose an input port to receive dynamic headers from another upstream component. This creates an issue with OAuth where the access_token expires every 10 minutes.

To bypass this limitation, I built a Python Custom Component in Langflow that handles both token renewal and tool invocation:

import requests

def _get_access_token(self) -> str: payload = { "grant_type": "refresh_token", "refresh_token": self.refresh_token, "client_id": self.client_id, "client_secret": self.client_secret, } response = requests.post(self.token_url, data=payload) response.raise_for_status() return response.json()["access_token"] This component runs on each execution, retrieves a Bearer token, calls the MCP endpoint with sql_exec_tool, and returns a StructuredTool for the Agent.

My questions are:

Is there an existing community component or built-in mechanism in Langflow to pass dynamic headers to the native MCP Tools component?

What is the recommended pattern in Langflow for connecting to MCP servers requiring short-lived Bearer tokens?

Environment: Langflow 1.11, Python 3.12, Snowflake native MCP.