Felippe Felisola Caso and Francesco Galletta present Monzo's internal agentAn AI coding agent is a tool-using AI system that can inspect, modify, and validate software within a repository. as an extension of existing engineering infrastructure, rather than a model working alone. Slack requests, task assignment and code review build on the bank's monitoring, staging environments, deployment tools and service integrations. The presenters report substantial internal adoption, but the talk's transferable lesson is how those building blocks reinforce one another.
Felippe Felisola Caso and Francesco Galletta describe an MCP gatewayAn AI agent gateway is a shared control layer that mediates agent access to models, tools, data and observability services. that exposes selected internal tools and knowledge while managing authentication and authorization. Remote agents run inside isolated namespacesAn AI agent sandbox is an isolated execution environment that limits which files, processes, networks, credentials, and external systems an agent can access. with explicit tools and controlled network access. This separates nondeterministic agent execution from production infrastructure while retaining useful context.
Felippe Felisola Caso and Francesco Galletta show applications in incident investigation, internal question answering and data-quality checks. Non-specialists can draft changes through the same entry point, but engineering review and staging remain part of deployment. The platform is designed to swap models and extend use cases without treating generated output as automatically production-ready.
Watch on YouTube
