MCP, tool calling, kill switches, human-in-the-loop and eval: agents that act instead of chatting, with clear limits and no disasters.
MCP, tool calling, kill switches, human-in-the-loop and eval: agents that act instead of chatting, with clear limits and no disasters.
How to design a kill switch for agents that write to Salesforce: payload dry-run, approval queue with signed links, t...
Prompt injection in company documents: how a PDF can hijack an AI agent toward an IBAN change or a double payment, an...
It isn't a tool war: it's a decision framework. n8n, LangGraph and plain Python on the same real case (booking + CRM ...
A chatbot that 'reads opportunities' closes nothing. To execute actions on the CRM you need an MCP tool layer designe...
Antonio Trento — System Architect & AI Integrator
I design and build data, backend, interface and AI agents end-to-end. No slides: systems that run and stay yours.
If a process costs you hours, clients or margin, it can probably be built. Let's talk.
Book a call