n8n-Claude Agentic Flow Setup

I’m wiring up n8n with Claude to create a hybrid “agentic” coworker that can both run on its own and pause for guidance when needed. The core of the job is a clean JSON workflow for n8n plus the Python glue code that lets Claude decide when to launch or chain my existing automation scripts. Scope • Build a single, import-ready n8n workflow (JSON) that starts from a manual trigger, routes the user prompt to Claude, and returns Claude’s structured response. • Add decision logic so the flow can either continue autonomously or branch for user confirmation—Combination of both autonomy and user input is essential. • In the autonomous path, call a Python node or webhook that runs predefined scripts (the main task is Running automated scripts). Claude’s answer should map clearly to the script name, parameters, and any environment variables. • Return execution results to Claude for a short summary, then post the summary back into n8n. Deliverables 1. n8n JSON file with all nodes, credentials placeholders, and comments. 2. One or two example Python scripts plus a lightweight runner/handler. 3. README that explains setup, Claude API keys, where to drop additional scripts, and how to extend the decision logic. Acceptance criteria • JSON imports into a fresh n8n instance without errors. • “Autonomous” branch executes the sample script end-to-end; output is visible in n8n. • “User input” branch gracefully waits for manual confirmation before proceeding. • Claude’s responses are parsed and used correctly for both branches. Mention any library choices you feel are best (e.g., FastAPI, requests, LangChain) and keep the code tidy and well-commented so I can expand it later.

N° de projet : 40271909