The context
A self-hosted n8n instance running close to two hundred workflows: WhatsApp and Telegram bots, API synchronisation between systems, scheduled jobs, and operational tooling such as deploy and log management.
What I did
- 1
Self-hosted on Docker
Runs on infrastructure I control, triggered by webhooks and schedules, so cost and data stay predictable.
- 2
Ops tooling the team actually uses
A git-push to production deploy that validates the secret and branch and rejects unauthorised calls, and a form-driven log manager that fetches or clears server logs over SSH and emails them as attachments.
- 3
Agentic pipelines, not just triggers
A daily workflow where an AI agent writes a script and captions as validated JSON, HeyGen renders the videos, and the result is published to YouTube, Shorts, TikTok, Instagram, Facebook and X, with a dedicated error path.
- 4
Built once, used for many clients
A single social-posting engine serves many brands with per-client configuration across 10+ platform branches. The largest single workflow is a 100+ node content pipeline from Google Sheets to multi-platform publishing and analytics.
The results
196 workflows self-hosted and running about 940 production runs a week, including the deploy, log and social-publishing automations above.




