NanoClaw is a self-hosted personal AI agent you message from the apps already on your phone. It suits individuals who want a private assistant inside WhatsApp, Telegram, Slack, or Discord without handing their life to a hosted service. This review is based on public official sources observed in October 2026, not hands-on testing. Version 2026.10.0 was current at observation with a commit one day prior, and a v1-to-v2 migration path documents an actively maintained project.[1][2][3][4][5]
Quick verdict
People who want a private assistant living inside their messaging apps will get the most from NanoClaw. Its per-session Docker containers, credential gateway, and per-agent memory keep each conversation isolated and personal. Anyone unwilling to run Docker, keep a computer on, and pay a separate AI provider should choose something else, since NanoClaw is self-hosted and brings no model of its own. If you want an assistant you can read in an afternoon and run yourself, choose NanoClaw; if you want zero setup, choose a hosted assistant instead.[1][2][3]
Key features
Thirteen messaging channels WhatsApp, Telegram, Discord, Slack, Microsoft Teams, iMessage, Matrix, Google Chat, Webex, Linear, GitHub, WeChat, and email via Resend connect through install-on-demand channel skills, and several can run at once.[1]
Per-session Docker isolation Each active session gets its own Linux container on macOS, Linux, or Windows via WSL2, so shell commands run inside the sandbox and agents see only explicitly mounted folders.[1][2]
Credential gateway Agents never hold raw API keys; outbound requests route through an installed gateway that injects credentials at request time and enforces per-agent policies and rate limits.[1][2]
Flexible isolation modes Each channel can map to its own private agent, share one agent with unified memory but separate conversations, or fold several channels into a single shared session.[1][2]
Per-agent workspace Every agent group keeps its own instructions file, memory, credentials, and mounts, so nothing crosses between agents unless explicitly wired.[1][2]
Scheduled tasks Recurring jobs run on the configured provider and message results back, with optional script gates that skip waking the agent when there is no work.[1][2][3]
Skills over features Channels and providers ship as installable skills from registry branches, so each fork contains only the integrations its owner chose.[1]
Provider choice per agent Claude via the official Agent SDK is the default, with Codex, OpenCode-backed providers, and local Ollama models as drop-in options configurable per group.[1][2]
Who will love it
Privacy-minded individuals gain a verifiable story for handing an agent personal conversations through the container boundary, credential gateway, and a codebase small enough to read.[1][2]
Messaging-first users get an assistant where they already talk, including scheduled briefings delivered to their phone.[1][2]
Tinkerers who fork benefit from the skills model and per-agent workspaces that reward trimming a fork to exactly their channels, providers, and tools.[1]
Who should avoid it
Anyone wanting zero setup should pick a hosted assistant, since Docker, a powered-on host computer, and a separate AI account are mandatory.[1][2]
Non-technical households face installer failures that hand off to Claude Code for diagnosis, which assumes terminal comfort.[1]
Teams needing managed guarantees find only a community Discord, with no enterprise support plan, SSO evidence, or managed hosting observed.[1][2]
Alternatives and compatibility
Questions worth asking
What is the best NanoClaw alternative for non-technical users?
Is NanoClaw really no-code?
How much does NanoClaw cost per month?
Sources and editorial review
Source research. Reviewed on October 11, 2026. Written by Sanjib Sah. AI assisted; reviewed by a human editor.
We have not run or tested this product.
- NanoClaw on OpenAlternative
official · Observed October 11, 2026
Discovery lead: official-site pointer, repository, MIT license, self-hosted flag, version 2026.10.0, commit one day before observation. Read October 11, 2026.
- NanoClaw: Secure AI Agent for WhatsApp, Telegram and More
official · Observed October 11, 2026
Official homepage: container isolation, OpenClaw comparison, channels, workspace, scheduler, skills, credential security, stars, funding banner. Read October 11, 2026.
- nanocoai/nanoclaw README
official · Observed October 11, 2026
README: motivation, quickstart, channel list, skills model, providers, requirements, architecture, migration. Read October 11, 2026. Star counts attributed to the catalogue snapshot, not this page.
- nanocoai/nanoclaw LICENSE
official · Observed October 11, 2026
Verbatim MIT License text, copyright 2026 Gavriel. Read October 11, 2026.
- nanocoai/nanoclaw package.json
official · Observed October 11, 2026
Package metadata: name, version 2026.10.0, Node >= 22, pnpm 10, TypeScript, better-sqlite3, ncl CLI. Read October 11, 2026.
- NanoClaw documentation: What is NanoClaw?
official · Observed October 11, 2026
Docs intro: self-hosted definition, requirements, AI accounts, memory, tasks, templates, migration. Read October 11, 2026.
- nanocoai/nanoclaw on GitHub
official · Observed October 11, 2026
Repository page confirming nanocoai/nanoclaw identity; identity evidence only. Read October 11, 2026.