AgentSandbox on by defaultTracked every 6 hours

NanoClaw

Small TypeScript agent (MIT) built on Anthropic's Claude Agent SDK. Every agent runs in its own Docker container, on macOS, Linux or Windows through WSL2.

GitHub stars
31,071
GitHub, 2026-11-02
Last release
v2026.10.0
2026-10-09
CVEs (30 days / total)
0 / 10
from NVD
Sandbox default
yes
GitHub stars, last 12 snapshots: +13

Overview

NanoClaw is a small TypeScript assistant from the nanocoai project, MIT-licensed since its first commit in January 2026. It is its own codebase, not a fork: its README says it gives OpenClaw's core functionality in a codebase small enough to understand. Each agent group runs in its own Linux container, with Docker as the runtime on macOS, Linux and Windows through WSL2, and only sees the folders mounted into it; API keys stay in a credential gateway outside the container. Setup is git clone and bash nanoclaw.sh, which installs Node, pnpm and Docker if they are missing and pairs a first channel (Slack, Telegram, Discord, WhatsApp, iMessage or a local CLI). Claude runs through Anthropic's Claude Agent SDK by default; OpenAI Codex, OpenCode (OpenRouter, Google, DeepSeek and others) and local Ollama models are skills you add, per agent group. Customisation means changing the code, with Claude Code. Until October 2026 this page called NanoClaw a macOS-only community fork installed from a .pkg and sandboxed in Apple containers; none of that matches its repository.

Our take

Editorial opinion, not a test result.

Best for

  • A personal assistant on chat apps where every agent runs in its own container
  • Claude users who are happy to customise the code itself with Claude Code
  • macOS, Linux or Windows (WSL2) hosts that already run Docker

Not for

  • Hosts without Docker: containers are how it isolates agents
  • Teams that want configuration files: NanoClaw changes are code changes
  • Setups that need agents cut off from the internet without extra work: the network lockdown is off by default

Security posture

Sandbox by defaultyes
Auth on dashboardyes
Threat model documentedyes
Multi-LLM supportyes
CVEs on NVD (total)10
CVEs in the last 30 days0

On by default: every agent group runs in its own Linux container (Docker on macOS, Linux and WSL2) as an unprivileged user with all capabilities dropped, and sees only the folders mounted into it. API keys stay in a credential gateway outside the container. The network lockdown that cuts containers off from the internet is off by default.

Sources

Sandbox default and model support read in the project's own documentation and repository on 2026-10-18.

Links

Run NanoClaw on portable hardware

Verified working on these portable hosts (with the standard caveats: see the per-device pages for power, RAM headroom, browser-tool support):

Intel NUC 13 and Geekom IT13
€300-600 · 15-55 W
Mac Mini M4 / M4 Pro
€699-2200 · 4-65 W (M4), 5-140 W (M4 Pro), idle to maximum per Apple
Mac Studio M3 Ultra
€4,500-7,000 · Up to 480 W (maximum continuous)
MacBook Air M3 / M4
€1,299-2,099 · 10-25 W

Compare NanoClaw with others

NanoClaw vs OpenClaw
No sandbox by default
NanoClaw vs Hermes Agent
No sandbox by default
NanoClaw vs Nanobot
No sandbox by default

NanoClaw CVEs (10 on NVD)

All CVEs


Stats refreshed every 6 hours from public sources (GitHub API, NVD CVE feed, CISA KEV). See our methodology for the full tracking pipeline.

See also: all AI agents, all AI CVEs, AI hardware, scan your AI stack, Pro alerts.