OpenClaw just dropped v2026.3.28. It’s the kind of release that makes you realize how fast this ecosystem is moving.
I updated my VPS yesterday. Here’s what I found.
🛡️ Plugin Approval Hooks
This is the sleeper feature. Any tool can now pause and ask for your OK before executing.
Running agents autonomously? Now you can have a human-in-the-loop for sensitive actions. Skills can prompt, you approve, the agent continues.
🤖 xAI Responses API + x_search
OpenClaw now integrates with xAI’s Responses API and x_search. If you’re running agents that need to fetch current information, this is a big deal.
Combined with the new approval hooks, you can have an agent search, present findings, and wait for your go-ahead before acting. Perfect for research workflows.
📱 Channel Fixes
- Discord & iMessage: ACP bind here — cleaner session handling
- WhatsApp: Echo loop fixed (no more agent talking to itself)
- Telegram: Message splitting improved for long texts
- Discord reconnect: More reliable after network blips
⚙️ How to Update
If you’re running OpenClaw on a VPS or local machine:
🦞 Why This Matters for StackOfTruths
I run 6 OpenClaw agents 24/7 on a VPS. They handle marketing reports, lead tracking, and content drafts. The new approval hooks mean I can let them do more without worrying about runaway actions.
And with Security Sentinel scanning every skill, the whole stack stays safe. Local-first, API-bill-free, and now with better control over what agents can actually do.
🚀 What’s Next
Paul (@thdxr) is in Tokyo for ClawCon. This release feels like a prelude to something bigger. The plugin system is getting serious. The gateway is solid. And the community is actually building useful things.
If you’re running AI agents, this is the release to upgrade for.
🦞 Stacking truths daily. One release at a time.
Leave a Reply