Skip to main content
Carry is mock-first: real adapters sit behind interfaces, and the system falls back to deterministic mocks when a key is absent — so it runs offline and a demo never hard-fails on a flaky testnet. Here’s exactly what’s real.

​
On the demo specifically

Seed memories

Real Walrus testnet blobs, uploaded once with apps/web/scripts/seed-walrus.mjs. Resolvable on any testnet aggregator.

New captures

Hit Walrus live and get a real blob id back.

Cross-model answers

Live GPT-4o and Claude calls when keys are present.

The revoke

A real policy change the gate honors before the model is invoked.
The honesty here is the point. Carry’s whole pitch is provability — so the docs say exactly what’s live, what’s mock, and the single thing that doesn’t persist, rather than pretend otherwise.