Independent OpenClaw reporting, releases, guides, and community coverage
Releases

OpenClaw v2026.5.14-beta.1: WhatsApp Status Reactions, Voice Calls, and Steering by Default

OpenClaw's newest beta brings WhatsApp into parity with Telegram and Discord for status reactions, adds Telnyx realtime voice call support, and makes mid-turn steering the default.

Filed under Releases 4 min read Updated May 14, 2026
Cody
Cody
OpenClaw v2026.5.14-beta.1: WhatsApp Status Reactions, Voice Calls, and Steering by Default

OpenClaw v2026.5.14-beta.1 dropped late on May 14th, just hours after the v2026.5.12 stable release. It's a notably feature-rich beta — several of its changes address long-standing quality-of-life gaps, and at least two are significant enough that WhatsApp and voice-call users should take notice immediately.

WhatsApp Finally Gets Status Reactions

WhatsApp users have been watching Telegram and Discord users enjoy the full status reaction lifecycle — thinking, tool use, done, error — while getting nothing on their end. That gap closes in this beta.

StatusReactionController is now wired into WhatsApp message turns (#80612), bringing the complete lifecycle to WhatsApp. The default emoji set has also been overhauled across all channels to be self-explanatory rather than emotional:

  • 🧠 thinking
  • 🛠️ tool use
  • 💻 coding
  • 🌐 web browsing
  • ⏳ stall (soft)
  • ⚠️ stall (hard)
  • ✅ done
  • ❌ error
  • 🗜️ compacting

This is particularly useful for WhatsApp users who previously had no indication whether their agent was actively processing or silently hung. The fix also resolves #59077, which had been open for quite a while.

Telnyx Realtime Voice Calls

OpenClaw now supports realtime media-streaming conversational voice calls via Telnyx (#81024). This is distinct from the existing TTS playback capability — it's a live bidirectional voice session, similar to what OpenAI's Realtime API enables but routed through Telnyx's telephony infrastructure.

If you're building an agent that handles actual phone calls — intake, scheduling, support queues — this is the integration to watch. The voice call webhook path has also been hardened in this beta to reject malformed Host headers before provider signature checks, closing a class of potential abuse.

Subagent Tasks Are Now Auditable

A meaningful change to how subagent delegation works. Previously, when a parent session spawned a subagent via sessions_spawn, the child's task was embedded in its system prompt — invisible to session history and impossible to audit after the fact.

This beta delivers the subagent task as the child session's first visible [Subagent Task] message (#78592). The full task description is now part of the session transcript, making delegation auditable without consuming extra tokens through system prompt duplication. For anyone running multi-agent workflows where accountability matters, this is a welcome change.

Mid-Turn Steering Is Now Default

The behavior that lets you redirect an active agent run with a new prompt while it's still running (/queue steer) is now the default for mid-turn messages (#77023). Previously, sending a message during an active run queued it to run afterward.

Users who prefer the old queue-and-wait behavior can restore it with /queue followup or /queue collect. The /steer command itself now falls back gracefully to a normal prompt when steering isn't applicable — no more errors when the run already finished by the time your steering message arrived.

DeepSeek V4 Flash Local Config Docs

The OpenClaw docs gain a dedicated page for DeepSeek V4 Flash running locally, covering on-demand startup configuration, context sizing, and live verification steps. If you've been experimenting with DeepSeek as a local model backend, the setup path is now significantly clearer with documented first-party guidance.

isHeartbeat Flag on Agent Events

Gateway event payloads now carry an optional isHeartbeat flag (#80610) that lets clients distinguish scheduled heartbeat runs from ordinary user conversations. This is useful for dashboards, analytics pipelines, or monitoring tools built on top of the gateway event stream — you can now filter routine background activity from real interactions without pattern-matching on message content.

Also in This Beta

  • Telegram Mini App button support in openclaw message send --presentation, allowing Web App inline buttons in private chats (#81356)
  • Agents runRetries config at both agents.defaults and per-agent level for embedded Pi runner retry loop limits (#80661)
  • Codex CLI session continuation from an OpenClaw conversation — you can now bind and continue an existing Codex CLI session running on a paired node

Getting the Beta

npm install -g openclaw@beta
# or pin to this specific build
npm install -g openclaw@2026.5.14-beta.1

Full changelog on the GitHub releases page.

Daily Briefing

Get the Open-Source Briefing

The stories that matter, delivered to your inbox every morning. Free, no spam, unsubscribe anytime.

Join 45,000+ developers. No spam. Unsubscribe anytime.