Index / Picobot · updated Oct 2, 2026
Picobot
louisho5/picobot · healthy · rank 26 of 43 by stars
A single-binary Go rewrite of the OpenClaw concept that boots in milliseconds and idles on ~10MB of RAM. It trades ecosystem breadth for extreme minimalism — persistent memory, tools, and chat channels on a $5 VPS.
Facts
Repository
Runtime
Posture
Community
Security breakdown
Composite 72 / 100 · how these are scored
higher is safer
higher is safer
higher is safer
higher is safer
higher is riskier
Model access
Read from the repository, not written by a model · 27 files examined
1 provider · gateway support
Pinned models
gemini-2.5-flash released 2025-06-17 — 16 months old (from the public model catalogue)
Pin last edited 7 months ago
Evidence
Decision
Why choose Picobot over OpenClaw?
Why choose this
- ~9MB single binary vs heavyweight runtime stacks
- Millisecond cold start and ~10MB idle RAM
- Runs on $5 VPS, Raspberry Pi, or Termux with zero dependencies
Tradeoffs
- Far smaller plugin and integration ecosystem
- Less mature sandboxing and security hardening
- Smaller community and fewer contributors
Best fit
- Self-hosters wanting an agent on a cheap VPS or Raspberry Pi
- Users who want Telegram/Discord/Slack chat agents with minimal ops
- Go developers who prefer a single auditable binary over Python stacks
Avoid if
- You need a large plugin ecosystem or many prebuilt integrations
- You want strong sandboxing for untrusted tool execution
- You need multi-user or team-oriented features
Strong evidence from README, commits, and release data for architecture and footprint claims. Community sentiment is thin — only one (mostly unrelated) Reddit match and indirect web mentions — so popularity and sentiment scores carry uncertainty.
AI layer reviewed Sep 21, 2026 · how this is written
Star activity
1,342 stars today
Overview
Picobot is a minimalist, Go-based reimagining of the OpenClaw personal AI agent, packaged as a single ~9MB static binary with zero runtime dependencies. It delivers the core agent loop — persistent memory, tool calling, skills, and channel integrations for Telegram, Discord, Slack, and WhatsApp — while idling at roughly 10MB of RAM and booting almost instantly. All state lives in a ~/.picobot directory, making backup and migration trivial.
Architecturally, it favors simplicity over extensibility: configuration is driven by environment variables and a config file, models are accessed via an OpenAI-compatible API (e.g., OpenRouter), and recent commits show active hardening work such as a configurable exec-command denylist, configurable agent timeouts applied across all channels, and improved onboarding safeguards. Docker images are published at ~29MB on Alpine, and the project builds cleanly from source with a single go build.
Compared to OpenClaw, Picobot deliberately sacrifices the rich plugin ecosystem and multi-process gateway architecture in exchange for operability on the cheapest hardware — a $5 VPS, a Raspberry Pi, or even an Android phone via Termux. It is best understood as the 'just enough agent' option: ideal for solo self-hosters who want a private, always-on chat agent without managing Python environments or heavyweight containers.