NanoClaw

Name: NanoClaw
Availability: OnlineOnly
Author: Qwibit

Autonomous Systems

A lightweight, container-isolated AI assistant that connects to WhatsApp with memory, scheduled jobs, and agent swarms via Anthropic's Agents SDK.

Visit Website

At a Glance

Pricing

Open Source

MIT-licensed open-source project, free to use and modify

Engagement

0views

0saves

0discussions

Available On

macOS

Linux

SDK

Resources

Website Docs GitHub llms.txt

Topics

Autonomous Systems Conversational Agents Multi-agent Systems

About NanoClaw

NanoClaw is a lightweight, open-source personal AI assistant built as a secure alternative to Clawdbot / OpenClaw. It runs inside isolated Linux containers on macOS or Linux, addressing the security concerns of larger agent frameworks by reducing the core logic to roughly 500 lines of TypeScript — making the entire system auditable in minutes rather than days.

NanoClaw connects to WhatsApp via the Baileys library and uses Anthropic's Claude Agent SDK to process messages inside isolated containers. Each chat group gets its own sandboxed environment with separate memory and filesystem access, preventing data leakage between contexts.

Container Isolation - Agents run inside Apple Container (macOS) or Docker (Linux) with filesystem-level sandboxing, so commands execute in the container rather than on the host machine
WhatsApp Integration - Send and receive messages through WhatsApp, including a private main channel for admin control and isolated group channels for different contexts
Agent Swarms - Spin up teams of specialized agents that collaborate on complex tasks within a single chat, with each sub-agent maintaining its own memory context
Scheduled Tasks - Configure recurring jobs that run Claude and message results back, such as daily pipeline briefings or weekly git history reviews
Per-Group Memory - Each group maintains its own CLAUDE.md memory file and isolated filesystem, keeping conversations and data separate
Web Access - Agents can search and fetch web content to incorporate real-time information into responses
Skills-Based Extension - Instead of adding features to the codebase, contributors create Claude Code skill files that teach the AI how to transform installations, keeping each fork minimal and purpose-built
AI-Native Setup - No configuration files or installation wizards; Claude Code guides the entire setup process including dependencies, authentication, and container configuration
Single-Process Architecture - One Node.js process manages a per-group message queue with concurrency control, using SQLite for persistence and filesystem-based IPC
Optional Integrations - Add Gmail, Telegram, Slack, and other services through community-contributed skills that rewrite the local codebase to fit exact needs

Community Discussions

Be the first to start a conversation about NanoClaw

Share your experience with NanoClaw, ask questions, or help others learn from your insights.

Pricing

OPEN SOURCE

Open Source

MIT-licensed open-source project, free to use and modify

Full source code access
Container-isolated agent execution
WhatsApp integration
Agent swarms support
Scheduled tasks

View official pricing

Capabilities

Key Features

Container-isolated agent execution via Apple Container or Docker
WhatsApp messaging integration with Baileys library
Agent swarms with parallel specialized agents
Scheduled recurring tasks with Claude execution
Per-group isolated memory and filesystem context
Web search and content fetching
Skills-based extensibility via Claude Code skill files
AI-native setup and configuration through Claude Code
Single-process Node.js architecture with SQLite persistence
Per-group message queue with concurrency control
Filesystem-based IPC for container communication
Optional Gmail integration via skills
Main channel admin control for managing groups and tasks
MIT licensed open-source codebase

Integrations

Anthropic Claude Agent SDK

Claude Code

Apple Container

Docker

SQLite

Gmail (via skill)

Obsidian (via mounted directories)

Reviews & Ratings

No ratings yet

Be the first to rate NanoClaw and help others make informed decisions.

Developer

Qwibit

Qwibit is an AI-native go-to-market agency for B2B tech companies, founded by brothers Gavriel and Lazer Cohen. Based in Israel, the company builds custom AI systems and uses them to help clients accelerate their sales and marketing operations. Qwibit grew out of the founders' experience running Concrete Media, an established B2B tech PR agency that Lazer founded in 2014 (originally as WestRay Communications). The brothers began building AI tools to automate their own agency workflows, and the results were strong enough to formalize into a dedicated venture. Gavriel Cohen brings the engineering side. Originally from New York City, he served as an IDF combat soldier before emigrating to Israel in 2013. He earned a BSc with honors in Computer Science and Physics from Tel Aviv University, then spent six to seven years as a software developer and team leader at Wix.com. He also serves as Vice President of Concrete Media. Lazer Cohen brings the media and business development side. Also originally from New York, he co-founded a PR firm there in 2009 before moving to Israel and launching Concrete Media. He has over a decade of relationships with top tech reporters and a deep understanding of how B2B tech companies get coverage and generate demand. Qwibit's most visible creation is [NanoClaw](/tools/nanoclaw/), an open-source personal AI assistant released on January 31, 2026 under the MIT License. NanoClaw was built as a secure, minimal alternative to OpenClaw, whose codebase had grown to roughly 400,000 lines with hundreds of dependencies. NanoClaw takes the opposite approach — about 500 lines of core TypeScript that can be fully audited in eight minutes. It runs AI agents inside isolated Linux containers using Apple Containers on macOS or Docker on Linux, so a compromised agent can only access resources explicitly mounted for it. The project is built on Anthropic's Claude Agent SDK and connects to WhatsApp for messaging. It gained over 7,000 GitHub stars in its first week. Internally, Qwibit runs a NanoClaw instance called "Andy" that manages the company's sales pipeline. Andy delivers daily briefings on lead statuses, assigns tasks to the team, and parses messy WhatsApp notes and email threads into structured updates stored in an Obsidian vault and SQLite database. It also handles recurring technical jobs like reviewing git history for documentation drift. This eat-your-own-cooking approach is central to how Qwibit positions itself — the agency doesn't just recommend AI tools, it builds and depends on them daily. Concrete Media, the sister company, remains active as a global B2B tech PR agency with a team of roughly ten professionals based primarily in Tel Aviv and Jerusalem. Its client base spans AI, machine learning, cybersecurity, developer tools, and MedTech, with coverage in outlets like VentureBeat and TechCrunch. The PR expertise feeds directly into Qwibit's GTM work, giving the agency an unusual combination of hands-on AI engineering and media fluency that most GTM agencies lack.

Founded 2025

Tel Aviv, Israel

Unknown (no pub... raised

10 employees