n8n AI Agent Workflow: CNAPS MCP Image Upscale 4x

Upscale images 4x with an n8n AI Agent connected to CNAPS.ai MCP—models chosen on the fly

This n8n AI Agent workflow sends your image (via URL) to CNAPS.ai through the MCP server, lets the agent select the best media model from 69 AI models, runs a complete processing chain, and returns the finished upscaled 4x image back in chat.

What this workflow does

• Chat request trigger: The n8nn-nodes-langchainchat trigger receives your prompt (e.g., “upscale this image 4x”) along with a file URL.
• Claude interprets the task: The agent uses n8nn-nodes-langchainlm chat anthropic to understand the intent and determine the required media operation.
• Calls CNAPS.ai via MCP: The workflow uses n8nn-nodes-langchainmcp client tool to:
  • run suggest_flow to choose a model,
  • use upload_file to register the input,
  • use create_flow to build the processing chain.
• Waits for completion: The agent polls with get_flow until the run finishes.
• Returns the output URL: The workflow posts the resulting upscaled file URL back into the chat.

Use cases

• Automation for creators: upscale product photos 4x before publishing.
• SaaS ops: improve image quality in bulk using a single chat endpoint.
• Engineering workflows: reuse the same agent flow as CNAPS.ai adds new models—no workflow rewiring needed.

Technical details

• CNAPS.ai MCP endpoint: https://mcp.cnaps.ai/mcp
• Server transport: HTTP Streamable (CNAPS.ai has no SSE endpoint, so SSE won’t connect)
• Authentication: Header Auth with credential name x-api-key; value is your CNAPS.ai API key (ensure allowed domains include mcp.cnaps.ai)
• n8n nodes/tools used: n8nn-nodes-langchainagent, n8nn-nodes-langchainchat trigger, n8nn-nodes-langchainmcp client tool, n8nn-nodes-langchainlm chat anthropic (plus a sticky note for context)