Affordable parts
Built around a Raspberry Pi, common servos, printable components, and an actively evolving bill of materials.
Molty is a four-legged robot pet you assemble yourself. It is being built to walk, talk, remember, and react when your AI agents get something done.
Walking now. Voice, memory, and agent connections are next.

REAL HARDWARE / JULY 2026
Molty is being built in public from printable parts, common servos, and a Raspberry Pi. Every stable step, failed gait, loose cable, and working integration becomes part of the open build.
Watch the latest 48-second demo01 / BUILD
Print the parts, connect the servos, and bring Molty to life. Because you build the body yourself, the relationship starts before Molty takes its first step.
Built around a Raspberry Pi, common servos, printable components, and an actively evolving bill of materials.
Replace parts, modify behaviors, and make your Molty different from every other one.
The BOM, failed gaits, fixes, and working demos are shared as they happen.
02 / BOND
Molty is being designed to greet you, learn routines, remember interactions, and express what it is doing through movement and sound.
No notification dashboard. No face trapped behind another screen. Just a pet that shares your desk.
The current prototype walks, waves, and dances with Raspberry Pi-powered motion control.
These companion behaviors are the next build stage, not capabilities of today's walking prototype.
03 / CONNECT
Connect Molty to the agents already working for you. When an agent finishes a task, finds an opportunity, or needs attention, Molty can react in the room.
Agent connections are in early experiments. The reaction below is a concrete behavior the project is building toward.
See the agent reactionCONCRETE EXAMPLE
Trading agent makes a profitMolty celebrates.
LIVE PROTOTYPE PROGRESS
Molty wobbles. Servos disagree. Cables escape. Every failure is public—and every fix moves the pet closer to becoming something people can build and live with.

HOW MOLTY WORKS
Molty does not wait for an AI model before moving. Balance and motion run through a fast local reflex loop. Planning, memory, and agent decisions happen through a slower reasoning loop.
That keeps the body responsive while the mind is still thinking.
Local motion control is working now, while gait refinement and recovery remain in development. The full agent loop is planned.
FAST LOOP
Live sensor data becomes motor commands for gait, balance, recovery, and responsive movement.
SLOW LOOP
AI models interpret context, plan actions, form memories, and decide what Molty should do next.
LONG-TERM GOAL
Better movement locally, and better decisions through memory and planning.
PRODUCT ROADMAP
Molty is a long-running robotics project, but the next steps are concrete: make the kit reproducible, make the pet expressive, and make agent connections useful.
01 / BUILD
Affordable parts, printable components, documented assembly, stable walking, and repairable hardware.
02 / BOND
Voice, memory, routines, and expressive movement grounded in what Molty can sense and do.
03 / CONNECT
Turn meaningful agent events into movement, sound, attention, and shared rituals.
BUILD UPDATES
Get the BOM, CAD files, and important build updates as they become available.
Follow @moltypet for build releasesUseful updates only. No weekly newsletter.
CREATOR OF MOLTY
“I'm building the robot pet I wanted on my desk: open, expressive, repairable, and connected to the agents already working for me.”