Agent substrates: the missing layer for AI agents

Tencent just shipped AI-Infra-Guard, a toolkit for keeping autonomous agents from doing whatever they feel like. The core idea is simple — agents need guardrails before they get sent off to browse the web or hit APIs. Without them, you're asking a language model to operate at scale and hoping for the best.

The name comes from the substrate pattern: an agent sits on top of a thin layer that constrains what it can touch. Think of it like a sandbox — the agent can read, it can write, it can call specific endpoints, but it can't reach the database directly or hit the billing API. You define the surface area, the substrate enforces it.

This is what happens when people building agent frameworks realize their first production incident was a model reading customer data it shouldn't have seen. The pattern is becoming a real thing in the infra layer — not a product so much as a discipline. You're writing guardrails the way you write rate limits: before you ship, not after.

Why this matters for us: the agents we're told will save us hours are just people with worse permissions — and the ones running side hustles need to lock theirs down before they burn a client.