Langflow MCP Server

What it does

Langflow MCP Server lets AI assistants directly manipulate Langflow visual flow graphs through conversation. Instead of dragging components on a canvas, you describe what you want — "add an OpenAI model and connect it to the output" — and the MCP server applies targeted differential operations to your flow in real time. It connects to a Langflow instance (local or the hosted graceful-ai.com platform) and exposes flow management, component discovery, and an AI assistant (Hopper) as MCP tools.

Tools

• tweak_flow — Apply differential operations (add/update/remove nodes and edges) to an existing flow without replacing it
• search_components — Search the 100+ Langflow component catalog by keyword
• get_flow — Retrieve the current state of a flow
• list_flows — List all flows in the connected Langflow instance
• create_flow — Create a new flow from scratch or a template
• chat_with_hopper — Multi-turn Claude-powered assistant that interprets natural language and applies flow operations
• undo / redo — Navigate flow history with full state tracking
• get_templates — Retrieve pre-built flow templates for common use cases

Installation

Clone and build the repo, then configure Claude Desktop:

{
  "mcpServers": {
    "langflow": {
      "command": "node",
      "args": ["/absolute/path/to/langflow-mcp/dist/api/mcp-server.js"],
      "env": {
        "LANGFLOW_API_URL": "https://www.graceful-ai.com",
        "LANGFLOW_API_KEY": "your-api-key-here",
        "ANTHROPIC_API_KEY": "sk-ant-xxxxx"
      }
    }
  }
}

Or run as an HTTP server for REST/SSE access:

npm run build && npm start

Supported hosts

Confirmed in README: Claude Desktop.