b stack radar
N

NanoClaw

6.4/10

NanoClaw is an open-source, MIT-licensed AI assistant that runs Claude-based agents in isolated containers. It connects to messaging apps such as WhatsApp, Slack, and Teams.

Data last updated November 2026 · Our methodology

Overview

NanoClaw is an open-source personal and team AI assistant built on Anthropic's Claude Agent SDK. It began as a deliberately small, security-focused alternative to OpenClaw. Each agent or group gets its own Linux container through Docker or Apple Container, with its own workspace, memory file, and filesystem. Agents can only see directories that are explicitly mounted. Outbound requests go through a credential gateway, so agents never hold raw API keys. It runs as a single Node.js process with SQLite state and supports scheduled recurring tasks and web search. You reach it through chat channels including WhatsApp, Telegram, Discord, Slack, Microsoft Teams, iMessage, Matrix, Google Chat, Webex, Linear, GitHub, WeChat, and email, each added on demand with a skill. You customize it by changing code in your own fork rather than through configuration files. Besides Claude, it can use OpenAI Codex, OpenCode via OpenRouter, Cursor, and Ollama. The software is free under the MIT license. Your real costs are the host machine and LLM API usage. NanoCo, the Tel Aviv company behind the project, raised a 12 million dollar seed round in May 2026 and sells enterprise assistant deployments, either self-hosted or vendor-hosted, but has not published prices. It suits developers and technical teams who want an auditable agent they can host themselves. Its main limitation is that it is built for hands-on users: setup happens in a terminal through Claude Code, observability is basic logs rather than a dashboard, and its ecosystem is smaller than OpenClaw's.

Our Rating

6.4
out of 10
Integration
7
Analytics
2
Customization
8
Ease of Use
4
Support
4
Cost
9
Scalability
5
Security
8
Output Quality
7

Who NanoClaw is best for

Best software for Developers

Good fit

  • Setup through a terminal and Claude Code and customization by editing code make it fit developers best.

  • Container isolation, a credential gateway, and self-hosting appeal to security-minded IT teams.

  • It is free to run on existing hardware and covers many chat channels.

Less suited

  • NanoCo's new enterprise offering adds auditing and private deployment, but it is young and unproven.

  • Non-technical teams will struggle with container setup and code-based customization.

Pros & Cons

Pros

  • Each agent runs in its own container and only sees mounted directories, which limits the damage a misbehaving agent can do.
  • The codebase is small enough for a team to audit, unlike OpenClaw's much larger platform.
  • It connects to more than a dozen chat channels, including Slack, Teams, WhatsApp, and email.
  • The software is free under the MIT license, so you pay only for hosting and model usage.
  • A credential gateway keeps raw API keys out of the agents.

Cons

  • Setup requires a terminal, Claude Code, and a container runtime, which is a barrier for non-technical users.
  • Observability is limited to logs, with no polished dashboard.
  • Its ecosystem of skills and guides is smaller than OpenClaw's.
  • It is built around Claude tooling, though other providers have been added.
  • The commercial vendor was founded in 2026, so its enterprise offering has a short track record.

Key Features

AI Agents
Private Deployment
Workflow Automation
Text Generation
Code Generation

NanoClaw Pricing

Open Source

Free
  • MIT license
  • Container-isolated agents
  • LLM API costs paid separately
  • Messaging channels added via skills
  • Self-hosted on macOS, Linux, or WSL2
Start Free

Enterprise (NanoCo)

Contact for pricing
  • Contact sales
  • Local model support
  • Agent actions auditable
  • Human approval for sensitive actions
  • Self-hosted or vendor-hosted deployment
Get Started
Open interactive pricing calculatorGet a tailored cost breakdown for your team size, billing cycle, and add‑ons.

Frequently Asked Questions

Yes. The software is MIT-licensed and free. You pay for your own hardware or VPS and your LLM API usage. NanoCo's enterprise deployments are sold through sales, with no public price.

Agents run in Docker or Apple Container Linux containers and can only access explicitly mounted directories. Credentials are handled by an external gateway rather than given to the agent.

WhatsApp, Telegram, Discord, Slack, Microsoft Teams, iMessage, Matrix, Google Chat, Webex, Linear, GitHub, WeChat, and email via Resend. Each is added through a skill.

Yes. Besides Claude, the README lists OpenAI Codex, OpenCode via OpenRouter, Cursor, and Ollama for local models.

NanoCo, founded in 2026 by brothers Gavriel and Lazer Cohen and based in Tel Aviv. It raised a 12 million dollar seed round in May 2026.

Ranked near NanoClaw in AI Software

See NanoClaw in the list

NanoClaw is #355 in AI Software by our score.

  1. #352
    Sonix logo
    Sonix
  2. #353
    Quickbase logo
    Quickbase
  3. #354
    Lovable logo
    Lovable
  4. #356
    Harvey logo
    Harvey
  5. #357
    GUIDEcx logo
    GUIDEcx
  6. #358
    DeepSeek logo
    DeepSeek