OpenClaw 2026.3.22 Release Adds Plugin Marketplace and Multi-Model Support
Openclaw

OpenClaw has announced its 2026.3.22 release, introducing major upgrades that push the platform further into becoming a full-fledged AI agent operating system rather than just a tool. 

The update brings a plugin marketplace, new model integrations, and improved agent reasoning, signaling rapid momentum in the open-source AI agent ecosystem.

ClawHub: A Plugin Marketplace for AI Agents

The headline feature in this release is ClawHub, a new plugin marketplace for OpenClaw.

This allows developers to:

  • Discover and install plugins

  • Extend agent capabilities

  • Share reusable tools and workflows

👉 This is a big shift — OpenClaw is moving toward an ecosystem model, similar to app stores or plugin marketplaces in IDEs.

Support for Latest AI Models

The release also adds support for newer AI models, including:

  • MiniMax M2.7

  • GPT-5.4 Mini and Nano

This means OpenClaw agents can now run on faster and cheaper models, improving scalability for real-world automation tasks. 

Per-Agent Reasoning Improvements

Another important upgrade is per-agent reasoning, which allows:

  • Different agents to use different reasoning strategies

  • Better task specialization

  • More efficient multi-agent workflows

👉 This aligns with the industry trend of multi-agent systems, where smaller agents collaborate instead of one large model doing everything.

“/btw” Side Questions Feature

The release also introduces a lighter feature called “/btw” side questions, enabling users to:

  • Ask quick follow-up questions

  • Maintain context without interrupting workflows

While small, it improves usability for day-to-day interactions with agents.

This update shows how fast OpenClaw is evolving:

1. From Tool → Platform

With ClawHub, OpenClaw is becoming a developer ecosystem, not just software.

2. Embracing Multi-Model Strategy

Support for models like GPT-5.4 Mini/Nano reflects the shift toward cost-efficient AI pipelines.

3. Agent Specialization

Per-agent reasoning hints at more advanced automation systems, where different agents handle different tasks.

Source: OpenClaw