Open sourceTypeScript · ≈31k stars · MIT
NanoClaw
Lightweight, container-isolated OpenClaw alternative built on Anthropic's Agents SDK.
NanoClaw is a personal AI assistant that runs each agent in its own Linux container, so agents get filesystem isolation instead of relying only on permission checks. Its author built it as a small alternative to OpenClaw, which the README describes as nearly half a million lines of code. NanoClaw is one process and a handful of files. It connects to WhatsApp, Telegram, Slack, Discord, Gmail and other channels, and supports memory and scheduled jobs. It runs on Anthropic's Agents SDK. Setup installs Docker, a credential gateway and the agent container. v2 can give each agent its own Slack bot identity, container and memory.
Pros & cons
Each point cites its sources and says what kind of evidence backs it. A second reviewer tried to refute every claim before it was kept.
Pros
- DocsPitched as an auditable alternative to OpenClaw, which its README describes as nearly half a million lines: NanoClaw is one process and a handful of files.[1][2]
- DocsEach agent runs in its own Linux container that sees only mounted paths, and a credential gateway injects API keys at request time so agents never hold raw keys.[1]
- DocsWide channel coverage (WhatsApp, Telegram, Slack, Teams, iMessage, Matrix, email via Resend and more), each added on demand with an /add-<channel> skill.[1]
- DocsMoved from Apple Containers to Docker, so it now runs on macOS, Linux and WSL2.[1][2]
Cons
- DocsClaude Code is the native runtime; Codex, OpenCode and Cursor are add-on skills that copy provider modules into your fork, after users pushed for non-Anthropic runtimes over subscription-ban fears.[1][2]
- DocsYou customize it by editing code in your own fork, and moving from v1 to v2 needs migrate-v2.sh, which hands off to Claude Code to replay fork customizations. Users criticized the incompatible branches that leave Claude to resolve merge conflicts.[1][2]
- User reportsHN commenters point out that container isolation does not stop data leaks or abuse once the agent is wired into email and chat.[1][2]
- User reportsOne early user said it broke down on a real task and saved credentials as JSON files in the repo (February 2026, v1).[1]
- Best for
- Individuals who want an always-on, chat-reachable Claude agent with container isolation and a codebase small enough to read.
- Watch out for
- Containers limit host damage but not what a connected agent can leak; keep mounts and channel permissions minimal.
Assessed Oct 29, 2026 · adversarially reviewedSuggest a correction
Neighbours on the workbench
TypeScript · ≈393k stars · MIT
openclaw/openclaw
The viral open-source personal AI assistant that really does things.
Maintained by OpenClaw Foundation
TypeScript · ≈265k stars · MIT
deepseek-ai/deepseek-harness
Everything is a Plugin — DeepSeek's open agent harness.
Maintained by DeepSeek
TypeScript · ≈258k stars · MIT
NousResearch/hermes-agent
The agent that grows with you.
Maintained by Nous Research
Connections
Relations from the Harness Hub knowledge graph, as charted in the State of Harness reports.
Maintained by · 1
- nanocoaiOrganisation
Built on · 1
- Claude Agent SDKProject
Similar entries · 1
- ZeroClawProject
Sources
Verified Oct 29, 2026