Your agent may pass the Turing test. But does it pass the Mom test?
Setting up a permissioned AI agent for a parent who isn't technical.

A few months ago I wanted to onboard my mom into OpenClaw.
The problem was not her. The problem was everything else.
To give someone access to an AI agent you hand them an API key in plain text, pray they don’t paste it somewhere, and hope the agent doesn’t decide to read emails it wasn’t supposed to. There is no security layer. The agent either has access or it doesn’t. You configure it in files you
edit by hand, in a terminal, with no guardrails on what it can actually reach.
I didn’t feel comfortable giving that to anyone.
So I started thinking about what it would take to do this right.
MCP as a security layer.
The Model Context Protocol gives you explicit, declarative control over what an agent can touch. Email, calendar, Drive — each one is a permission you grant or withhold. The agent operates within that boundary. It cannot self-escalate. It cannot reach outside its
allowed scope. You define the surface, the agent works within it.
Out-of-band permissions.
The permission model lives outside the agent runtime. The agent doesn’t control its own access. You do, through a separate layer that the agent cannot influence. This matters because an agent that can modify its own permissions is an agent you cannot trust.
Containerized runtime.
Each agent runs in its own isolated container. One agent misbehaving does not affect another. You can shut it down without touching anything else. The blast radius of any failure is contained by design.
A UI a non-developer can actually use. No terminal. No config files. No YAML. Onboarding is a Telegram conversation. You answer a few questions, connect your services, and the agent is running. That’s it.
Today I onboarded my mom. She has a personal AI agent handling her email and calendar, and sending her reminders for high priority items. She has not asked me what a config file is. I consider this a success.
The platform is called AgentHelm. The domain is agenthelm.mom.
As it turns out, literally.
I’m looking for a small group of beta testers. The ask is small: a MiniMax API key, connecting your Gmail to what is, admittedly, a suspicious-looking app, and telling me what breaks.
Originally published on Medium