OpenClaw lands TTS upgrade, OpenTelemetry coverage, browser hardening (2026.4.25)
TL;DR
OpenClaw 2026.4.25 ships a TTS upgrade with /tts latest, chat-scoped auto-TTS, six new voice providers, OpenTelemetry coverage across the agent loop, browser automation hardening, and PWA / Web Push support.
What changed
OpenClaw 2026.4.25 (April 27, 2026) lands a full TTS upgrade with /tts latest, chat-scoped auto-TTS controls, personas, and per-agent / per-account overrides, plus new provider coverage for Azure Speech, Xiaomi, Local CLI, Inworld, Volcengine, and ElevenLabs v3. Plugin startup and install paths move to the cold persisted registry, cutting broad manifest scans.
OpenTelemetry coverage expands across model calls, token usage, tool loops, harness runs, exec processes, outbound delivery, context assembly, and memory pressure with bounded low-cardinality attributes.
Browser automation gains safer tab URLs, iframe-aware role snapshots, CDP readiness tuning, headless one-shot launch, and deeper browser doctor probes. Control UI and setup flows add PWA / Web Push support, Crestodian first-run repair, TUI setup, context mode selection, and a shorter startup greeting. Install hardening covers Windows, macOS, Linux, Docker, bundled plugin runtime deps, Node service restarts, LaunchAgent token rotation, and mixed-version gateway verification.
Why it matters
The OpenTelemetry coverage is the production-readiness move. Standard agent frameworks emit traces only for outbound HTTP; instrumenting the entire loop (tool calls, exec processes, context assembly, memory pressure) is what lets you actually debug slow agent runs. Bounded low-cardinality attributes mean the trace volume stays manageable at scale. The TTS expansion to six new providers collapses voice-product integration cost dramatically.
What to watch for
If you run any agent in production, wire up the new OTel exporter and validate that your APM backend (Tempo, Honeycomb, Datadog, New Relic) ingests the bounded attribute set without cardinality explosions. If you ship voice features, benchmark Azure Speech vs ElevenLabs v3 vs Inworld on your actual audio prompts before standardizing on one provider. Skim the install hardening notes for anything affecting your deployment platform.
Who this matters for
- Developers: Adopt 2026.4.25 if you run voice features (six new TTS providers including Azure, ElevenLabs v3, Inworld), OpenTelemetry-instrument the agent loop, or care about install hardening across Windows / macOS / Linux / Docker.
Harsh’s take
The OpenTelemetry coverage is the operational win. Most agent frameworks emit traces only for HTTP calls; OpenClaw 2026.4.25 instruments model calls, token usage, tool loops, harness runs, exec processes, outbound delivery, context assembly, and memory pressure with bounded low-cardinality attributes. That is the difference between knowing an agent run was slow and knowing exactly which step in the loop burned the latency budget.
The TTS expansion is good for voice-first products: Azure Speech, Xiaomi, Local CLI, Inworld, Volcengine, and ElevenLabs v3 all land as first-class providers with chat-scoped auto-TTS and personas. If you are building a voice agent, this collapses provider integration cost. The browser automation hardening (safer tab URLs, iframe-aware role snapshots, CDP readiness tuning) is the kind of fix that matters when you scale headless runs.
by Harsh Desai
About OpenClaw
View the full OpenClaw page →All OpenClaw updatesGo deeper
More AI news
- Daily RoundupAuK and Edge0 models trend on Hugging Face, GPT Image 2.5 on Fal, plus Altman delays IPO
On 12 September new speech and text models hit Hugging Face while OpenAI's image tool launched on Fal, and industry commentary stressed human skill over raw AI output even as Altman postponed an IPO.
- Daily RoundupOpenAI gpt-image-2.5 models hit Replicate and Fal, Tailscale ships model router, and safety warnings rise
OpenAI released two new image models on Replicate and Fal while Tailscale and Featured showed production AI use on Vercel; safety and governance debates intensified across labs.
- Weekly DigestCursor Origin hosting and agent swarms, Claude Code fixes, Codex GPT-6-Astra (agent tools ready now)
Cursor rolled out native code hosting and multi-agent orchestration while Claude Code and OpenAI Codex shipped targeted fixes and new model access for terminal workflows.