n8n-workflow-patterns

Use the pattern Webhook → Validate → Transform → Action → Response. Remember that incoming payload data is nested under $json.body, not $json directly, and choose a response mode: onReceived for immediate 200 responses or lastNode for custom responses.

Connect a language model to the AI Agent node via the ai_languageModel port, attach any node as a tool via the ai_tool port, and add Window Buffer Memory via ai_memory for conversation context. Write specific tool descriptions since the AI uses them to decide when to call each tool.

Use Schedule → Query → Transform → Write → Verify with incremental sync queries filtering on updated_at timestamps. Always use parameterized queries to prevent SQL injection, batch large datasets with Split In Batches, and use transactions for multi-step writes.

Webhook payload data is nested under $json.body, so {{$json.email}} returns nothing while {{$json.body.email}} works. Headers are under $json.headers, URL parameters under $json.params, and query strings under $json.query.

Use a loop pattern with offset-based, cursor-based, or Link-header pagination depending on the API. Initialize a page or cursor variable, check the response for a next indicator, increment, and loop back until no more pages exist.

Add an execution lock using Redis: check for a lock key at workflow start, skip execution if it exists, otherwise set the lock with a TTL, run the workflow, and delete the lock on completion. Also set the workflow timezone explicitly to handle DST correctly.