Ai

OpenClaw v2026.4.15 Adds Gemini TTS, Auth Health Visibility, and Remote Memory Storage

OpenClaw v2026.4.15 Adds Gemini TTS, Auth Health Visibility, and Remote Memory Storage

OpenClaw v2026.4.15 delivers a meaningful platform update focused on model operations, voice capabilities, memory infrastructure, and usability for teams running mixed provider environments. This release adds Gemini text-to-speech support, improves visibility into model authentication health, expands durable memory storage beyond local disk, and introduces a GitHub Copilot embedding path for memory search.

What Changed

Anthropic model defaults were updated. OpenClaw now shifts default Anthropic selections, opus aliases, Claude CLI defaults, and bundled image understanding toward Claude Opus 4.7. This helps standardize behavior across Anthropic-powered workflows and reduces ambiguity around which Claude tier is active by default.

Gemini text-to-speech support was added to the Google plugin. The bundled google plugin now supports Gemini TTS with provider registration, voice selection, WAV reply output, PCM telephony output, and setup guidance. This expands OpenClaw's native voice stack and makes it easier to support both conversational and telephony-style audio delivery inside one provider integration.

The Control UI now shows model auth health. A new Overview card exposes OAuth token health and provider rate-limit pressure, with clear callouts when tokens are expiring or already expired. Under the hood, this is powered by a new models.authStatus gateway method that removes credentials from responses and caches results for 60 seconds.

LanceDB memory can now use cloud storage. The memory-lancedb integration now supports remote object storage for durable memory indexes instead of requiring local disk only. That opens the door to more portable and resilient memory deployments, especially in distributed or server-based environments.

Memory search gained GitHub Copilot embeddings support. OpenClaw adds a GitHub Copilot embedding provider for memory search, plus a reusable Copilot embedding host helper for plugins. The transport layer is designed to respect remote overrides, token refresh, and safer payload validation, which should make multi-plugin reuse cleaner and more reliable.

Experimental lean mode was added for local models. A new setting, agents.defaults.experimental.localModelLean: true, can drop heavyweight default tools such as browser, cron, and message. The goal is to reduce prompt size for weaker local-model setups without changing the normal default path for standard deployments.

Why It Matters

This version is notable because it improves both the operator experience and the infrastructure flexibility of OpenClaw. The new auth status card addresses a common operational pain point by making token expiry and provider pressure visible before failures become disruptive.

The addition of Gemini TTS is also significant for teams building voice-enabled agents, customer support flows, or telephony integrations. Native support for both WAV and PCM-style outputs makes the feature practical beyond demo use cases.

On the memory side, remote object storage support for LanceDB is a meaningful architectural step. It allows durable memory to move beyond single-machine assumptions, which is especially useful for cloud-hosted agents and more scalable deployments.

Meanwhile, GitHub Copilot embeddings support gives developers another path for semantic memory search, while the new lean local-model mode shows continued attention to smaller, cheaper, and more constrained inference environments.

Overall, OpenClaw v2026.4.15 looks like a platform maturity release. Rather than one headline feature, it delivers a set of practical improvements that make deployments easier to monitor, voice features easier to ship, and memory infrastructure more adaptable.

Official Source: https://github.com/openclaw/openclaw/releases/tag/v2026.4.15