Hey Product Hunt - I'm Zorie, and we built The Membrane.

Why: teams are handing AI agents write access to production. Agents that merge code, change infrastructure, edit tickets and touch customer data. When one does the wrong thing, what's left is scattered logs and third-party chat histories. They can't prove an action was authorized by a policy that was still in force.

The Membrane is a small Rust gateway that sits in front of your agents:

- Every model and tool call needs a live, signed, time-bounded scope. No scope, no call.

- Each allowed action writes a tamper-evident, hash-linked receipt, so any action traces back to the exact policy and issuer that authorized it.

- A replayed proof, a silent model swap or an expired scope is refused. If the receipt chain breaks, the agent is severed.

It enforces before production is touched; it doesn't just observe after the fact.

Try it:

- Quickstart: clone to a refused call in about five minutes. No Docker, no paid API: https://github.com/membrane-hq/the-membrane

- Browser demo: https://demo.themembrane.org - a recorded replay of the gate flow; tool effects are simulated and labeled on screen

- Site: https://themembrane.org

Who it's for: teams letting agents touch production systems who need proof of authorization, not just logs.

Honest state: v0.1.0, early stage, single maintainer. The gate, signing and receipt chain are real and runnable today; the hosted demo's tool effects are simulated.

What we'd most like to hear: if you run agents with production access, what would you need to see before putting a gate like this in the path?