I press keys on my keyboard (mostly) in the correct order to make the machine do what I want.
My current work is centered around reducing friction: making it easier for humans and AI agents to access the right context, tools, and execution environments without turning every workflow into glue code.
Right now, I’m working on:
- a headless harness for running and observing AI-agent workflows in a predictable way
- Usher, a self-hosted credential broker that gives agents scoped, auditable access to the systems they need without spreading long-lived secrets
- Loom, a personal data lake platform for collecting, connecting, and making personal context useful across tools and agents
The common thread is infrastructure that gets out of the way: fewer manual handoffs, fewer brittle integrations, and better defaults for people building with agents.
- 🤖 Cave - A Cloud-first Agent Orchestrator that you can deploy on your own infrastructure (currently in private beta; will be open source soon).
- 🚪 Usher - A self-hosted credential broker for authenticated HTTP calls that keeps long-lived secrets out of prompts, logs, scripts, and model context.
- 🫰🏽 Generous - Platform that helps builders create a sustainable lifestyle.
- 🛫 Runway - A native macOS menu bar app that monitors GitHub Actions workflow.
- 🔁 filedrop - A simple file sharing service for AI agents.
- 🎛️ proxctl - A lightweight CLI for interacting with a Proxmox instance / cluster.
- X: @ItsAndreKoenig
- Website: https://andrekoenig.com
- Company: Open Formation GmbH
