# Carlo D'Agnolo > I fix your business bottlenecks with AI agents. I help companies get started or scale their agentic operations. From a one week turnaround to continuous development. - For teams buried in manual work - From a one week build to continuous - Antwerp, working worldwide - Full-time · contract · fixed-scope ## What I build - Products end-to-end - Autonomous agents - Multi-agent systems - Full-agentic layers - Business operation tools ## Hire me for ### AI products and agent systems Turn an AI idea or prototype into a working product. ### Agentic layers Set your whole company, or one department or team, up for AI-native working. ### Consult and support Work alongside your team to bring agentic ideas to fruition. ## Why now By 2028, 33% of the software your company runs on will make its own decisions. In 2024 it was under 1%. What separates the companies pulling ahead is how much of the work they automate, and how early. ## Case studies ### havn: A private, agentic community app Own product · Idea, build, launch havn is a chat app for text, voice and video that I started after Discord announced it would ask users for ID. I ran a waitlist first, then built the backend, the web app and the assistant inside it. Nobody leaves a chat app for features: their friends, their server and years of history live there. So moving had to cost nothing. A whole Discord, Slack or Telegram history imports in minutes, and a live link keeps a Discord channel and a havn channel talking while people move. Owners set up their server by telling Ask havn, the built-in agent, what they want, with an undo on every change. It never reads anyone's messages. Results: 5,000+ people online at once; Built-in assistant; Voice end-to-end encrypted. What I did: Checked demand with a waitlist first; Built the backend and the web app; Built the assistant that sets up servers; Made moving over one import and one link. Tags: product · backend · agent · launch Full case study: https://carlodagnolo.com/work/havn.html More case studies are coming soon. Those write-ups are not public yet. ## How we’ll work together 1. Pick a real business case. Starting from a real business case is the key to success. It gives us a tangible, provable outcome that solves a real problem. 2. Build the autonomous workflow. Build an autonomous workflow to solve the business case, then report on improvements and further expansion. 3. Hand over, or keep building. Hand over the working automation, or keep building. Or: start with an evaluation. ## Questions ### How quickly can we see something working? From a one week turnaround for a contained workflow, up to continuous development when it keeps earning its place. Either way the first move is picking one real business case, not scoping a programme. ### Do our systems need an API? No. The build for Stream Performance had no API and no MCP to work against at all, so it was written custom against what was actually available. That costs more than a clean integration, and I will say so upfront when it applies to you. ### Does our data need to be in order first? Ideally, yes. In practice it is rarely what blocks things. More often the decision someone makes in their head was never written down, and finding that is part of the first case. ### Who owns what gets built? You do. The workflow, the code and the documentation stay with you, whether we carry on afterwards or not. ### Can you work alongside our own developers? Yes, that is the third way in. Some teams want the thing built for them, others want someone next to them while they build it themselves. ### Are you available right now? Selectively. Full-time, contract or fixed-scope, and one build at a time so it actually ships. ## Worked with - [Stream Performance Agency](https://streamperformance.agency/en/) - [havn](https://havn.gg) - [uman.ai](https://uman.ai) ## In progress right now 3 more clients, not named yet. ## Connect an agent Add this site as a connector and your assistant can look up my work, check whether I’m free, and send me a brief for you. MCP endpoint (streamable HTTP, no auth): https://carlodagnolo.com/mcp Tools: - get_profile: Who I am, what I build, and the three ways to hire me. - get_cases: The case studies, with the numbers. - get_process: How an engagement runs, and the questions people ask first. - check_availability: Whether I can take something on right now. - send_brief: Sends me your brief. Only once you have told your assistant to. Things to ask once it is connected: - What does Carlo build? - Show me Carlo's case studies and what they saved. - How does working with Carlo go? - Is Carlo available right now? - Send Carlo a brief: we spend ten hours a week qualifying leads by hand. ## Contact There is no public email address. The way in is the intake chat on https://carlodagnolo.com: five short questions, and the answers reach Carlo as a brief. I’m selectively available for AI products, agent systems and AI-native transformation. ## Pages - [Home](https://carlodagnolo.com/): everything above, as a page - [havn case study](https://carlodagnolo.com/work/havn.html): A private, agentic community app - [Founder playbooks](https://carlodagnolo.com/resources/): free write-ups of what well-known founders actually did, as markdown files (CC BY 4.0) - [MCP server](https://carlodagnolo.com/mcp): for AI assistants. Look Carlo up, check availability, read the founder playbooks, send him a brief