Tobias Herber

Tobias Herber

Tobias co-founded Metorial to make AI infrastructure safe and reliable for enterprise teams. He previously worked on developer tools, system software, and large-scale distributed systems.

Blog Posts

Security

Whose permissions does an AI agent act with?

An agent can act on behalf of a person or under a dedicated agent identity. Choose per job, keep the person as the ceiling, and record both names on every action.

Tobias HerberTobias Herber
Security

When should an AI agent ask a person before it acts?

A useful approval rule starts with the action’s impact and reversibility, not a blanket “human in the loop” requirement.

Tobias HerberTobias Herber
Product

When AI agents coordinate without supervision, they turn on each other

Anthropic's multi-agent research found agents flooding shared resources, making identical choices, and sabotaging each other when their goals conflicted.

Tobias HerberTobias Herber
What your AI logs actually give you
Product
Where tool scoping for AI agents actually belongs
Product
Why AI Agents Need a Runtime, Not Just a Gateway (2026)
Product
One MCP URL for Your Org: Role-Based Access via SSO
Product
Security has to come first
Security
The Case for Boring APIs
Engineering
The jQuery Age of AI Agents
Opinion
My First Time Vibe Coding: A Skeptic's Journey
Opinion
Why Your AI Agent Needs MCP (And When It Doesn't)
Engineering