PicoClaw
AI agents on a $10 board
About
Runs on $10 RISC-V boards with less than 10MB of RAM. One-second boot time. The tiniest claw in the sea.
At a Glance
Runs On
LLM Required
Yes — needs API key
MCP Support
No
Ecosystem Role
Reimplementation
Performance
Startup: 1s
RAM: <10MB
Runs on $10 RISC-V boards, <10MB RAM, 1s boot
Editorial Take
PicoClaw sits in the OpenClaw family as a reimplementation rather than a generic AI agent tool. It is written in Go and currently shows 23k GitHub stars, which gives a quick sense of where it fits between experimentation and mainstream adoption.
PicoClaw is best understood as a response to the tradeoffs in OpenClaw itself. Compared with the flagship project, its community is roughly 12x smaller, but that smaller scope often exists because the project is optimizing for a sharper use case: minimal-footprint environments.
Operationally, this project targets minimal-footprint environments. Its deployment surface spans Embedded / IoT / RPi.
Best For
Minimal-footprint environments
- + A large GitHub footprint (23k stars) gives it more tutorials, contributors, and social proof than most projects in the ecosystem.
- + Its platform story explicitly includes constrained hardware, not just servers and laptops.
- + The published performance profile suggests a deliberate focus on runtime efficiency rather than feature sprawl.
Tradeoffs
What to understand before choosing PicoClaw over another project in the ecosystem.
- - It still depends on an external LLM provider, so token spend and vendor reliability remain part of the operating model.
- - Teams standardizing on MCP will need custom integration work or a different tool choice.
- - Its deployment focus is opinionated, so it may not fit teams that want a conventional server-first setup.
GitHub Stats
#4 of 41 on the leaderboard →Stars
★ 23k
+1.4k this week
Forks
3.0k
Open Issues
487
Last Commit
0d ago
What People Are Saying
Forget the Mac Mini: Run This OpenClaw Alternative for Just $10
PicoClaw ultra-lightweight personal AI Assistant runs on just 10MB of RAM
"Ditch your Mac Mini — now you can run a full AI assistant on $10 RISCV hardware with 10MB RAM."
OpenClaw Relationship
PicoClaw is an OpenClaw-inspired project — part of the growing ecosystem of personal AI agent frameworks that trace their lineage to Peter Steinberger's original self-hosted assistant.
Added to Shelldex on February 25, 2026
Compare with...
Similar Projects
ClawGo
Headless OpenClaw client in Go
ZClaw
Personal AI assistant on ESP32 — 888 KiB firmware
GoGogot
Lightweight self-hosted AI agent in Go
Golem
Your AI agent, single binary, zero dependencies
Nanobot
4,000 lines of pure research
ZeroClaw
Zero overhead, zero compromise