Ai

OpenClaw v2026.4.22 Expands xAI Media Tools and Local Terminal Chat Setup

OpenClaw v2026.4.22 Expands xAI Media Tools and Local Terminal Chat Setup

OpenClaw v2026.4.22 is a meaningful product update centered on multimodal AI operations, smoother local usage, and faster model administration. This release adds broad xAI media capabilities, expands realtime transcription support across more providers, introduces a local embedded terminal chat mode that works without a Gateway, improves first-run onboarding by auto-installing missing plugins, and makes it possible to register new models directly from chat without restarting the gateway.

What Changed

The biggest addition in v2026.4.22 is expanded xAI provider support. OpenClaw now supports xAI image generation, text-to-speech, and speech-to-text, including the grok-imagine-image and grok-imagine-image-pro models, reference-image editing workflows, six live xAI voices, multiple audio output formats such as MP3, WAV, PCM, and G.711, plus grok-stt transcription and xAI realtime transcription for Voice Call streaming.

Speech tooling also gets broader provider coverage beyond xAI. Voice Call streaming transcription is now available for Deepgram, ElevenLabs, and Mistral, extending the existing realtime paths that already supported OpenAI and xAI. ElevenLabs additionally gains Scribe v2 batch transcription for inbound media, which improves offline or asynchronous audio handling options.

For terminal users, the TUI now includes a local embedded mode. That allows teams to run terminal chats without a Gateway while still preserving plugin approval controls. For local-only workflows, this lowers setup friction without dropping safety gates around tool use.

Onboarding is also more resilient in this version. OpenClaw can now auto-install missing provider and channel plugins during setup, which removes a common first-run failure point where configuration could stall until plugins were manually recovered.

The OpenAI Responses integration has been refined as well. When web search is enabled and no managed provider is explicitly pinned, direct OpenAI Responses models now automatically use OpenAI's native web_search capability. If a managed provider such as Brave is pinned, OpenClaw keeps using the managed web_search tool instead.

Another practical admin improvement is the new /models add <provider> <modelId> command. Users can now register a model from chat and begin using it without restarting the gateway. The /models command remains available as a provider browser, but now includes clearer guidance and copy-friendly add examples.

The release also begins extending WhatsApp reply behavior with configurable native reply quoting through replyToMode, although the changelog excerpt provided here is truncated before the full details of that change are shown.

Why It Matters

This version strengthens OpenClaw as a serious multimodal operations layer. Native support for xAI image, voice, and transcription workflows means teams can route more media tasks through one orchestration surface instead of stitching together separate integrations. That is especially useful for assistants, voice interfaces, and content workflows that depend on both generation and transcription.

The broader streaming transcription support is also important for production voice systems. By adding Deepgram, ElevenLabs, and Mistral to the realtime mix, OpenClaw reduces provider lock-in and gives operators more flexibility to optimize for latency, cost, or model preference.

The local embedded TUI mode makes OpenClaw easier to use in developer and operator environments where a full Gateway is unnecessary or undesirable. Combined with preserved approval gates, it offers a more lightweight deployment path without abandoning core control mechanisms.

Operationally, auto-installing missing plugins during setup and allowing model registration from chat both reduce friction at critical moments. One improves first-run reliability, while the other speeds up experimentation and rollout when teams need to test or adopt newly available models quickly.

Official Source: https://github.com/openclaw/openclaw/releases/tag/v2026.4.22