Table of Contents
If you’re reading this article, you probably already know what OpenClaw (formerly Clawdbot / Moltbot) is, or you’ve at least bumped into it while looking for a self-hosted, tool-using AI assistant you can actually control.
That’s the key point: OpenClaw isn’t “just” a chat UI. It’s a platform you run on your own infrastructure that can connect to the channels you already use (messaging apps, web chat, automations), call tools, schedule reminders, and keep long-lived context in a way that feels closer to a real assistant than a one-off chatbot prompt.
At the center of that ecosystem there’s the OpenClaw Gateway: the control plane that exposes APIs (including WebSocket) to send messages into a session and receive assistant replies back. Once you have a Gateway, you can build lightweight clients around it. And that’s where ClawTalk comes in.
What is ClawTalk?
ClawTalk is a small Chrome extension (Manifest V3) that lives in the Chrome Side Panel and turns your OpenClaw Gateway into an always-available “talk to my assistant” interface.
The concept is deliberately simple:
- Keep a side panel open while you browse.
- Press Connect to attach to your Gateway.
- Press Talk, speak naturally, and pause.
- The transcript is sent to OpenClaw (into a specific
sessionKey). - The assistant’s reply appears in the panel—and can be spoken aloud using TTS.
The project is here: https://github.com/Ryadel/ClawTalk
Why build this if OpenClaw already has a web UI?
Because a classic chat tab is not the same as a voice loop.
When you’re using OpenClaw as an assistant, you might want to do something else while chatting with it. Maybe you’re already doing something else in the browser, you don’t want to switch tabs, you don’t want a heavy UI always visible, or you want a fast “press-talk-release” rhythm that feels natural.
ClawTalk is designed around that: minimal friction, minimal UI, and enough robustness to not melt your browser if the network is flaky (bounded buffers, sensible reconnect behavior, and conservative protocol usage).
ClawTalk connects to the Gateway over WebSocket and sends your messages into OpenClaw using the standard event mechanism (the equivalent of a “send chat message to a session”). You pick a sessionKey (for example main), and ClawTalk routes everything there.
This matters because it means you can reuse the context you already have: if your web chat or Control UI is using the same session key, ClawTalk becomes a “voice remote control” for that same assistant brain.
Installation
ClawTalk can be installed in Chrome in developer mode:
- Open
chrome://extensions - Enable Developer mode
- Click Load unpacked
- Select the
ClawTalk/folder
After installation, open the extension and choose Side panel. The panel gives you Connect/Talk controls, a chat area, and Settings.
Configuration overview
Most configuration lives in the side panel Settings. The main categories are:
- Gateway: URL, token, and optional headers
- Session: session key you want to talk to
- Speech: language defaults for speech recognition
- TTS: SpeechSynthesis (default) or ElevenLabs (optional)
1) Gateway URL
If your Gateway runs locally, the common default is:
ws://127.0.0.1:18789
If your Gateway is remote, you’ll use its public WebSocket URL. In that case, be mindful of access control (token + network layer protections).
2) Gateway Token
Set the Gateway Token (or device token, depending on how your OpenClaw instance is secured). ClawTalk stores it locally in Chrome storage and avoids printing it in logs.
3) Cloudflare Zero Trust (Cloudflare Access) headers
If your Gateway is behind Cloudflare Zero Trust (Cloudflare Access), you can use ClawTalk by setting a service token and configure it in the ClawTalk settings panel. To do that, add these in the Gateway headers configuration:
CF-Access-Client-Id: YOUR_CLIENT_IDCF-Access-Client-Secret: YOUR_CLIENT_SECRET
These values come from your Cloudflare Zero Trust dashboard where you generate an access Service Token. Once configured, ClawTalk will attach them so Cloudflare allows the WebSocket to reach your Gateway.
IMPORTANT: treat these values like credentials. Don’t paste them into screenshots, don’t commit them, and rotate them if you suspect exposure.
4) Session key
Choose a sessionKey (default is usually main). If you want shared context with your existing OpenClaw chat UI, use the same session key there too.
Text-to-Speech: SpeechSynthesis vs ElevenLabs
ClawTalk supports two TTS paths, and the choice is mostly about quality, privacy, and dependency.
Option A: Default browser SpeechSynthesis
This is the simplest setup: it uses the voices available in your browser/OS.
- No external API calls for TTS (good for privacy and reliability)
- No API keys, no billing
- Voice quality depends on the system voices installed (sometimes great, sometimes robotic)
- Best when you want a “works everywhere, minimal setup” experience
Option B: ElevenLabs TTS (optional)
ElevenLabs is a cloud text‑to‑speech (TTS) and voice AI platform. Among other things, it provides an API that takes text (for example, an assistant’s reply) and returns natural‑sounding speech audio using high-quality voices. Developers use it when they want voices that usually sound more human than the typical built‑in system voices.
If you want ClawTalk to have higher-quality, more natural voices, you can use ElevenLabs as long as you configure it in Settings.
To do that, you’ll need a ElevenLabs API Key and a ElevenLabs Voice ID: both of them can be created from the EvelenLabs Developers dashboard.
Trade-offs:
- Audio generation happens via external requests to ElevenLabs
- Quality is often noticeably better (especially for long replies)
- Usage depends on your plan/limits and connectivity
ClawTalk includes a Test speech button so you can validate your TTS setup without running the whole talk loop.
Conclusion
ClawTalk exists because the “assistant experience” isn’t just about models: it’s about ergonomics. A side panel with a stable voice loop can make an OpenClaw session feel like an actual tool you keep at hand, not a tab you occasionally visit.
If you already run OpenClaw, ClawTalk is an easy next step: configure the Gateway, pick the right session key, add Cloudflare Access headers if needed, and choose between a zero-dependency TTS (SpeechSynthesis) or premium voice quality (ElevenLabs).
