I help founders and engineers ship real AI agent systems.
Head of Open Source at Agency AI Solutions, Lisbon.
Agent control plane
Every AI agent on my machine — Claude, Codex, Kimi, OpenCode — runs under one control plane I built.
- One database holds every task, escalation and session summary.
- Mandates set what each agent may touch, and an agent asks for more when it needs it.
- The operating system enforces those file permissions, scoped to what each task could damage.
- When an agent needs a decision, the escalation reaches me at the system level.
This is the problem I work on.
What I build and ship
-
Mobile-first padel decision reps that turn coach advice into match instinct.
-
Voice-first European Portuguese practice in short real conversation scenes.
- In development
A voice-first assistant, private by design, that runs entirely on your device.
Open source
-
Reliable multi-agent orchestration framework. Lead maintainer and v1 architect.
-
Claude Code for everything except coding: complete deliverables from one prompt.
-
Terminal app for running and testing Agency Swarm projects.
Background
-
Automation and systems integration — Deloitte Consulting, Kaspersky, Logifuture.
-
30+ enterprise agent deployments.
-
Architect and #1 contributor — 1,200+ commits.
Say hello
GitHub (opens in a new tab)X (opens in a new tab)LinkedIn (opens in a new tab)