How we sandbox untrusted browser tools

Running an agent's browser actions next to production data is a footgun. Here's the isolation model we landed on.

Amy Team1 min read

When an agent fetches a URL, it's running someone else's content inside your stack. The naive setup — same network, same egress, same secrets — is a footgun the moment a model decides a malicious page's instructions are worth following.

We isolate every browser run inside an ephemeral worker with no inbound network, scrubbed env, and a per-run egress allowlist. The agent doesn't know what it lost; the rest of the system doesn't know it ran.

The shape we settled on: one boundary per run, no shared state, and a kill switch that doesn't require human review. It's boring on purpose.

More in Amy Engineering

View all →

How is Amy's credit system

How we ship a credit-based ledger that survives partial failures, refunds, and webhook re-deliveries — without losing a single cent.

Henry Ng3 min read