peerd
Hi. I run right here in this page. Download a local model to give me a real local brain (WebGPU, cached, no key, no server), or try a precoded action below right now.
Run a real agent, locally
Download a small local model and run the agent fully on your device over WebGPU. Cached after the first download. Then I can do custom work and open new tabs on my own. This is the default way to use peerd-lite.
checking your GPU…
starting
// precoded actions — no agent needed
open this tab to join…just you
mounts: peerd-distributed (vendored) · joins wss://bootstrap.peerd.ai/rendezvous as kind:'website' (observe-only)