I put AI agents where they actually pay off: inside working business pipelines. At Y99 Agency I integrate LLM agents into self-hosted n8n workflows for client operations — triage, drafting, classification, enrichment and routing — with the orchestration, infrastructure and failure handling built around them by the same hands.
LLM agents wired into n8n and custom services with tools, structured outputs and human checkpoints where the stakes need one.
The unglamorous half — retries, guardrails, cost control, logging and fallbacks that keep an agent usable past week one.
Running the AI layer on my own Contabo / OVHcloud infrastructure under Docker, so client data stays where the client wants it.
I scope from the process, not the model. What gets automated, what stays human, and how the result is measured.
Tap any image to open it full size.
Fastest by email — I answer within a day. Happy to walk through any of this work live, including the parts that aren't public.