OpenClaw March 29 Release: MiniMax Image Gen, xAI Responses API, Async Plugin Approvals

The OpenClaw team shipped a packed minor release overnight, touching providers, plugins, security workflows, and CLI tooling in one sweep.

Headline Features

MiniMax Image Generation — OpenClaw now ships a bundled image generation provider for MiniMax’s image-01 model, supporting both generate and image-to-image editing with aspect ratio control. The model catalog has also been trimmed to M2.7 only, dropping legacy M2, M2.1, M2.5, and VL-01 models.

xAI Responses API + x_search — The bundled Grok provider moves to xAI’s Responses API, gains first-class x_search support, and auto-enables the xAI web search plugin from existing config — no manual toggling required. Onboarding flows have been updated to offer x_search setup during openclaw onboard and openclaw configure --section web.

Async Plugin Approval Hooks — A significant new plugin capability: requireApproval can now be added to before_tool_call hooks, letting plugins pause tool execution and prompt the user for approval. This works across the exec approval overlay, Telegram buttons, Discord interactions, and the /approve command on any channel. The /approve command has been unified to handle both exec and plugin approvals.

apply_patch Enabled by Default for OpenAI — OpenAI and OpenAI Codex models now have apply_patch enabled by default, with sandbox policy access aligned to write permissions.

Other Notable Changes

  • ACP current-conversation binds: Discord, BlueBubbles, and iMessage now support /acp spawn codex --bind here to turn the current chat into a Codex-backed workspace without creating a child thread.
  • CLI backends expand: Gemini CLI joins Claude CLI and Codex CLI as a bundled inference backend; gateway run --claude-cli-logs is replaced with the generic --cli-backend-logs.
  • Podman improvements: Container setup simplified for rootless users; the launch helper now installs to ~/.local/bin.
  • Matrix TTS: Auto-TTS replies now send as native Matrix voice bubbles instead of audio attachments.
  • Slack explicit file uploads: A new upload-file action routes uploads through Slack’s native upload transport with optional filename/title/comment overrides.
  • Config cleanup: Legacy speech config auto-migrates on read; old bundled TTS API key shapes no longer fall back in normal mode. Config migrations older than two months are now dropped automatically.
  • Qwen provider deprecated: The portal.qwen.ai OAuth integration is removed; users must migrate to Model Studio with --auth-choice modelstudio-api-key.

Fixes

  • Agents/Anthropic: unhandled provider stop reasons (e.g. sensitive) now return as structured assistant errors instead of crashing the agent run.
  • Google/Gemini 3.1: pro, flash, and flash-lite now resolve correctly across all Google provider aliases.
  • OpenAI Codex image tools: image analysis no longer fails on missing provider registration.
  • WhatsApp: fixed an infinite echo loop in self-chat DM mode.

Full changelog: openclaw/openclaw on GitHub

← Back to News