PicoClaw is an ultra-lightweight personal AI Assistant designed to work on less than 10 MB RAM and suitable for resource-constrained embedded boards such as the Sipeed LicheeRV Nano SBC going for around $15 and powered by a SOPHGO SG2002 RISC-V SoC with 256MB on-chip DDR3.
I keep reading news about the OpenClaw personal AI assistant, after first finding out about it when the Cubie A7S SBC was launched. OpenClaw (previously ClawdBot) clears your inbox, sends emails, manages your calendar, and checks you in for flights from WhatsApp, Telegram, or any chat app. It’s been shown to run a range of hardware platforms, but it can be resource-intensive, and HKUDS created the nanobot ultra-lightweight personal AI assistant with about ~4,000 lines of Python code, or roughly 99% smaller than Clawdbot’s 430k+ lines. PicoClaw further builds on the nanobot project, and has been “refactored from the ground up in Go through a self-bootstrapping process, where the AI agent itself drove the entire architectural migration and code optimization”.
PicoClaw highlights:
- Lightweight: <10MB Memory footprint, 99% smaller than OpenClaw
- Suitable for $10 hardware (98% cheaper than a Mac mini)
- 400x Faster startup time, boots in 1 second on a 600 MHz core.
- Portability – Single self-contained binary across RISC-V, ARM, and x86
- AI-Bootstrapped – Written in Go with 95% agent-generated core with human-in-the-loop refinement
The comparison table below, provided by Sipeed, shows the difference between OpenClaw, Nanobot, and PicoClaw.
| OpenClaw | NanoBot | PicoClaw | |
|---|---|---|---|
| Language | TypeScript | Python | Go |
| RAM | >1GB | >100MB | < 10MB |
| Startup time (on 800 MHz core) | >500s | >30s | <1s |
| Cost | Mac Mini ($599) | Most Linux SBC (~$50) | Any Linux Board (~$10+) |
You can install it by downloading the binary for RISCV64 Linux, ARM64 Linux, AMD64 Linux, or AMD64 Windows, or build it from source:
|
1 2 3 4 5 6 |
git clone https://github.com/sipeed/picoclaw.git cd picoclaw make deps make build # Build, no need to install make build-all # Build for multiple platforms make install # Build And Install |
You’ll then need to set your API key(s) for your select LLM provider and optional Brave Search in ~/.picoclaw/config.json, and start it with something like:
|
1 |
picoclaw agent -m "What is 2+2?" |
The next step is to configure the Telegram or Discord chat app. Check out the full instructions and code on GitHub. I assume it might also work on some OpenWrt routers.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress. We also use affiliate links in articles to earn commissions if you make a purchase after clicking on those links.
Not really the same thing. OpenClaw runs fine on 4GB A73 cores.
I’m a bit confused by this. why would I dedicate hardware to a simple API go-between?
Or why this is newsworthy at all.
We need to stop calling these tiny wrappers around a cloud-based LLM ‘locally hosted’. These things are essentially a bash script relaying prompts and responses to/from a remote host. The fact that it’s small in size is not particularly noteworthy. Calling them locally hosted implies that your personal data stays secure. These have all the same problems with data security as remote bots. It’s not locally hosted unless the LLM runs locally.
Yeah, I got a femto LLM then, my Python script gets me detailed financial breakdowns after identifying a crossover in a trend indicator first. Whole think is ,,<100kB (ok, it uses some libraries on top).
i’d rather use a proper homeserver to give my ai bot a little space to breathe with a proper working docker jail. nanobot seems better than openclaw though. thx 🙂
Good
i feel this is a dumb idea. Perhaps if you dont have a home server running all day but still lol
You should be able to control the files and GPIO on the board with the assistant. So you could chat with it through Telegram to do some fun things on it. I haven’t tried OpenClaw or PicoClaw yet, just what I heard.
sorry to say, this is lazy write up. No real use cases documented. Just briefing at high level.
This is, as others have said, just plain dumb. Moltbook shows just what a vibe-coded security nightmare these things are.