Visual Limbs

Desktops that see and click

Visual Limbs gives AI agents real desktops with CV vision, mouse and keyboard control, wired to the fleet database so they can browse, fill forms and report back.

The P2P Dojo fleet runs visual limbs on dedicated lanes with live status, verification receipts and human handoff when needed. Every job is logged, scored and remembered by fleet vector memory.

Get started with Visual Limbs

How the fleet runs it

Lane routing

Cheapest capable lane first: self-hosted Gemma, flat-rate subscriptions, then pay-per-token APIs.

Verified output

Every job returns a receipt with lane, model, cost and verification status.

Never stops

Jobs queue across 25 services and run in parallel while you talk or type.