nanocoai/nanoclaw

Containerized agent runtimes are not new, but most force you to either cobble together your own channel integrations or accept a bloated framework. This one takes a different approach: ship a lightweight container that already speaks WhatsApp, Telegram, Slack, Discord, and Gmail out of the box, then layer persistent memory and scheduled jobs on top. The security angle is genuine — running agents inside containers reduces blast radius if something misbehaves, which matters more than most founders admit when they are handing an LLM access to live communication channels. The Anthropic Agents SDK underpinning means you get structured tool use and multi-step reasoning without wiring it yourself. For someone building a customer support bot, an internal ops assistant, or a notification workflow across multiple channels, this cuts weeks off the plumbing work. The honest reservation: the project is early and lightly documented, so expect to read source code when things go wrong. Production reliability is unproven at scale. -> Best for: indie hackers and small-team founders shipping multi-channel AI assistants who want container security without framework bloat.