A tiny AI assistant in Go that runs on $10 Linux boards with about 10MB of RAM.
Running an assistant on very cheap, low-power hardware
Free and open source (MIT). You pay only for the models you call.
Run an assistant on a small board or camera at the edge of your network.
Keep a chat assistant running 24/7 without paying for a full server.
Pair it with a KVM device to automate routine server checks.
Close alternatives worth a look before you decide.
PicoClaw is an AI agent from Sipeed. It runs as self-hosted and is offered under a open source model.
PicoClaw is offered under a open source model. Model tokens are usually billed separately, so check the live model price on this page and the vendor site for current tiers.
The alternatives section on this page lists the closest options, with live data for each. Use the head-to-head pages to compare two of them in detail.
Core memory use under 10MB, with fast boot even on a single slow CPU core.
One binary for RISC-V, ARM, MIPS and x86 boards, desktops and Android phones.
Connect Telegram, Discord, WhatsApp, Slack, Matrix, LINE and more than 19 platforms in total.
Send simple questions to lighter models and harder ones to stronger models.