Upgrade Android Talk Mode to realtime voice sessions
TL;DR
Switched Android Talk Mode to use realtime Gateway relay voice sessions, featuring streaming mic input, realtime audio playback, tool-result bridging, and on-screen transcripts.
What changed
OpenClaw switched its Android Talk Mode to realtime Gateway relay voice sessions on 22 May 2026. The update adds streaming mic input, realtime audio playback, tool-result bridging, and on-screen transcripts.
The change replaces the prior async message flow with a direct relay that keeps voice and tool calls in sync.
Why it matters
Vibe Builders who run OpenClaw on a phone now get closer to live conversation instead of waiting for heartbeat replies. This narrows the gap with closed apps that already offer realtime voice, but keeps the self-hosted memory and messaging integrations intact.
The move pressures competitors that still treat mobile voice as a secondary add-on. It also raises the bar for any self-hosted agent that wants to stay competitive on daily use.
How to use it
Update to the latest OpenClaw release through the CLI on your host machine. Enable Android Talk Mode in the YAML config, then connect the mobile client to the Gateway relay endpoint.
The feature is available now for users on the current main branch. Test first on a low-cost VPS to watch token spend before moving to production workflows.
Watch for
Confirm the bet if transcripts stay accurate and tool calls fire without lag during longer sessions. The bet breaks if audio dropouts appear under real network conditions or if token costs rise faster than expected.
Expect a follow-up release that adds similar realtime support to at least one desktop messaging client next.
Who this matters for
- Vibe Builders: Deploy the new Gateway relay to turn your phone into a low-latency, voice-first personal agent.
- Developers: Integrate the realtime Gateway relay into your mobile forks to sync tool calls with audio streams.
Harsh’s take
OpenClaw moving to realtime voice sessions via a Gateway relay is a necessary pivot to match the low-latency expectations set by proprietary models. By ditching the asynchronous message flow, they are solving the awkward 'walkie-talkie' lag that plagues most self-hosted mobile agents. The inclusion of tool-result bridging within a live audio stream is the real win here: it allows for interactive debugging and task execution via voice without breaking the session state.
Operators should watch the token burn closely. Realtime streaming is notoriously expensive compared to batch processing. If you are running this on a high-end model, your API costs will spike.
The technical challenge now shifts from connectivity to cost management and ensuring the on-screen transcripts remain accurate during network handoffs. This update sets a high bar for open-source mobile agents, forcing competitors to move beyond simple STT/TTS wrappers.
by Harsh Desai
About OpenClaw
View the full OpenClaw page →All OpenClaw updatesMore from OpenClaw
- App UpdateAdd doctor warnings for hidden MCP server tools
The doctor utility now warns users when sandbox tool policies hide configured Model Context Protocol (MCP) server tools before provider requests are made.
- App UpdateExpose stale-running task maintenance details in CLI
The tasks maintenance CLI command now outputs detailed JSON explanations for retained and reconcile candidates, detailing backing-session, cron, CLI, and wedged-subagent states.
- FeatureSupport per-agent lean local-model configuration
Added support for the experimental localModelLean configuration option on individual agents. This allows lean local-model mode to be enabled selectively rather than globally.