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 between onReceived and lastNode response modes based on whether the caller needs a custom response.

Connect a language model to the ai_languageModel port, any node to the ai_tool port, and a memory node to ai_memory on the AI Agent node. Write specific tool descriptions since the AI uses them to decide when to call each tool, and use read-only database credentials for safety.

Use webhook processing for event-driven integrations, HTTP API integration for fetching external data, database operations for sync and ETL, AI agent workflows for conversational or tool-using AI, and scheduled tasks for recurring reports and maintenance.

Webhook payload fields live under $json.body, so {{$json.email}} returns nothing while {{$json.body.email}} works. Also verify the HTTP method matches the sender and that you are reading body, query, params, or headers correctly.

Loop with offset, cursor, or Link-header pagination until no next page exists, and add Wait nodes or exponential backoff between batches. Set continueOnFail to true and check rate limit headers to pause before limits are exceeded.

Always use parameterized queries with placeholders like $1 and a parameters array instead of interpolating expressions into the query string. Combine this with a least-privilege database user that has only the permissions the workflow needs.