OpenClaw just shipped the kind of release that makes the whole stack feel more real. Version 2026.3.28 adds async plugin approval hooks, xAI Responses API support, broader image and CLI options, and a pile of fixes across chat surfaces and model providers. The big takeaway is simple: OpenClaw is getting safer and more usable at the same time, which is exactly what serious adoption needs. https://github.com/openclaw/openclaw/releases

OpenClaw is breaking out of the builder bubble and into mainstream tech coverage. CNN now frames it as a major consumer-tech story in China, which means this is no longer just a terminal toy for agent diehards. Once a product becomes both a dev tool and a cultural story, the growth curve usually gets weird in a hurry. https://www.cnn.com/2026/03/29/business/china-openclaw-ai-anxiety-intl-hnk-dst

The ecosystem is racing to turn agents into real operators instead of clever copilots. Roy Osherove's Loki aims to let OpenClaw deploy full-stack apps across AWS accounts, while LarkSuite CLI opens up docs, calendars, meetings, and tables to agent workflows. That is the pattern to watch: the winning tools are the ones that expand what agents can touch in the real world. https://x.com/RoyOsherove/status/2038197740362948611 https://x.com/ring_hyacinth/status/2037741103013896543

The most interesting OpenClaw stories are coming from people shipping with it, not just talking about it. YuChen says they built a trading strategy script with OpenClaw despite not being a coder, then shared a frontend and profitability tests after spending about $70 on API calls. That is messy, expensive, and exactly the kind of experimental workflow that tells you a platform has real pull. https://x.com/YuChen/status/2038226774207889719

@openclaw, shipping the new release
"Version 2026.3.28 adds plugin approval hooks, xAI Responses API integration, and fixes across WhatsApp, Telegram, and Discord."
https://x.com/openclaw/status/2038084923517796839

@RoyOsherove, building for solo founders
"Loki lets OpenClaw build and deploy full-stack apps autonomously across AWS accounts."
https://x.com/RoyOsherove/status/2038197740362948611

@ring_hyacinth, wiring agents into the workplace stack
"OpenClaw agents can now access docs, knowledge bases, tables, meetings, and calendars through LarkSuite CLI."
https://x.com/ring_hyacinth/status/2037741103013896543

@HeardFloor, throwing cold water on the hype
"OpenClaw is way too unstable."
https://x.com/HeardFloor/status/2037774382962278727

LarkSuite CLI integration gets the nod this week because it is the clearest example of where agent tooling is headed. It plugs OpenClaw into the systems teams actually use to run work: docs, knowledge bases, calendars, meetings, and tables.

Why it is cool is that it moves agents closer to operational reality. An agent that can read your notes is interesting. An agent that can work across the same collaboration stack as your team is useful.

How to get it: start with the demo post, then grab the project from GitHub. https://x.com/ring_hyacinth/status/2037741103013896543 https://github.com/larksuite/cli

YuChen used OpenClaw to build a trading strategy script despite not being a coder, then shared a frontend and profitability tests after roughly $70 in API spend. That is a real person using an agent to cross the gap from idea to working system, with enough cost and output detail to feel honest instead of polished. https://x.com/YuChen/status/2038226774207889719

The takeaway: the best agent demos are starting to look a lot like scrappy businesses.

The traps are baited, the ropes are tight, and the lobster fleet is sailing into bigger water.

If this crustacean dispatch stops sparking joy, the unsubscribe link below will help you slip quietly back into the sea.