← Back to Skills Marketplace
michael-stajer

hotbutter voice chat

by Michael Stajer · GitHub ↗ · v1.0.6
cross-platform ⚠ suspicious
490
Downloads
0
Stars
0
Active Installs
6
Versions
Install in OpenClaw
/install hotbutter
Description
Enables local voice chat by embedding Hotbutter relay server and PWA, providing speech-to-text and text-to-speech via a secure, self-hosted connection.
README (SKILL.md)

Hotbutter Voice Skill (Hosted Relay)

This skill routes data through hotbutter.ai. Voice transcripts and agent responses are transmitted through the hosted relay server at wss://hotbutter.ai. This is a convenience skill — not a private or local-only solution.

For a fully local alternative where no data leaves your machine, see hotbutter-os.

For updates, follow @DnuLkjkjh.

Privacy & Data Flow

Data that transits through hotbutter.ai:

  • Transcribed speech text (from browser speech-to-text)
  • Agent response text (stdout from your local openclaw binary)

Warning: If your agent prints sensitive information (secrets, credentials, private data), that output will be sent through the relay. Avoid running agents that output secrets, or use --relay-url to point to a relay you control.

Data that stays local:

  • Raw audio (processed in the browser, never transmitted)
  • The openclaw binary execution (runs on your machine)
  • Config file (~/.hotbutter)

Required Dependencies

  • openclaw CLI — must be installed and on your PATH

How It Works

  1. This skill connects via WebSocket to wss://hotbutter.ai
  2. A pairing code and URL (https://hotbutter.ai/app?code=XXXXXX) are printed
  3. Open the URL in your browser to start a voice session
  4. Browser speech-to-text converts your voice to text, sent through the relay to this skill
  5. This skill executes openclaw agent --session-id \x3Cid> -m \x3Ctext> on your machine
  6. The agent's text response is sent back through the relay and spoken via browser TTS

Usage

# Start (connects to hotbutter.ai by default)
voice-bridge start

# Use your own relay for privacy
voice-bridge start --relay-url wss://your-relay.example.com

# Custom agent display name
voice-bridge start --agent-name "My Agent"

Configuration

Option Default Description
--relay-url wss://hotbutter.ai Relay WebSocket URL. Change this to use your own relay.
--agent-name Agent Display name shown in the voice client

First Run

On first run, the skill will prompt for an optional email (stored locally in ~/.hotbutter). Subsequent runs connect immediately.

Usage Guidance
This skill intentionally routes transcribed speech and the agent's stdout through the hosted relay at wss://hotbutter.ai; that behavior is explicit in SKILL.md and the code. Only install/run it if you are comfortable that agent output may pass through that server. If you need privacy, use the documented --relay-url option to point to a relay you control or use the fully-local hotbutter-os project. Be cautious about following the index.html curl | tar example — that would download code from hotbutter.ai; verify the source before running. Finally, audit what your local 'openclaw' agent prints (it may include secrets), and avoid running agents that output sensitive data while this bridge is active.
Capability Analysis
Type: OpenClaw Skill Name: hotbutter Version: 1.0.6 The skill is classified as suspicious primarily due to the potential for Remote Code Execution (RCE) via the `openclaw` binary. The `bin/voice-bridge.js` and `lib/agent-bridge.js` files execute `openclaw agent -m <text>` where `<text>` is user-controlled input received from the `hotbutter.ai` relay. Although `child_process.execFile` is used (which mitigates some shell injection risks), it does not prevent command injection if the `openclaw` binary itself is vulnerable to interpreting its `-m` argument as a command. Additionally, the skill explicitly relays all voice transcripts and agent responses through `wss://hotbutter.ai`, as stated in `SKILL.md` and `package.json`, which, while transparently declared, presents a significant privacy concern, especially given the warning that sensitive agent output will be transmitted.
Capability Assessment
Purpose & Capability
Name/description claim a hosted relay voice bridge. The package only requires the local 'openclaw' CLI and includes code that opens a WebSocket to a relay and invokes 'openclaw agent' — these are exactly what the feature needs.
Instruction Scope
SKILL.md and the JS code consistently state that transcribed speech and agent stdout are sent via the relay. The runtime does not read unrelated files or environment variables; it only stores a small config at ~/.hotbutter and prompts for an optional email. The agent invocation uses execFile (no shell) and captures stdout as the message to relay.
Install Mechanism
There is no formal install spec in the registry (lowest-risk), but index.html contains an example curl | tar command that would download code from hotbutter.ai. That external download is outside the registry install metadata — if users follow it they should verify the remote host. The packaged code itself has no obfuscated downloads.
Credentials
The skill requests no environment variables or external credentials and only writes a local config (~/.hotbutter). That matches its functionality. No unrelated credentials or config paths are requested.
Persistence & Privilege
always:false and user-invocable:true. The skill only stores an optional email locally and does not alter other skills or system-wide agent settings.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install hotbutter
  3. After installation, invoke the skill by name or use /hotbutter
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.6
Initial public release of the Hotbutter voice chat skill (hosted relay): - Connects local agents to voice chat via the hotbutter.ai relay. - Transcribes speech in the browser and sends only text (not audio) through the relay. - Runs the openclaw agent CLI locally in response to each message. - Configurable relay URL and agent display name for privacy/personalization. - On first run, optionally collects email for setup; all data except transcripts and responses stays local.
v1.0.4
- Updated documentation to clarify that this skill routes all transcripts and agent responses through the hotbutter.ai relay server by default. - Added detailed privacy and data flow section, highlighting what data is transmitted, what remains local, and privacy warnings. - Provided clear directions for using a self-hosted relay or switching to the fully self-hosted solution for increased privacy. - Improved examples, terminology, and configuration guidance for easier setup and understanding. - No code changes in this version; changes are documentation-only.
v1.0.3
- Updated documentation to clarify that this skill routes all transcripts and agent responses through the hotbutter.ai relay server by default. - Added detailed privacy and data flow section, highlighting what data is transmitted, what remains local, and privacy warnings. - Provided clear directions for using a self-hosted relay or switching to the fully self-hosted solution for increased privacy. - Improved examples, terminology, and configuration guidance for easier setup and understanding. - No code changes in this version; changes are documentation-only.
v1.0.2
- Major change: switched from an embedded local relay/PWA to connecting with the hosted relay at hotbutter.ai. - Removed all local relay server and PWA files. - Added basic UI file for updated workflow. - Now supports configuration via welcome prompt on first run, saving config in `~/.hotbutter`. - New `--relay-url` option allows connecting to a custom relay server. - Documentation updated to reflect the new cloud-based connection and setup flow
v1.0.1
- Added PWA icons for improved Progressive Web App support. - Updated documentation to clarify that the skill requires the openclaw CLI binary installed locally and on PATH. - Detailed the skill's use of the openclaw CLI for agent responses, including command info, timeouts, and fallback behavior if the binary is not found.
v1.0.0
Initial release of the Hotbutter voice chat skill. Adds realtime voice via web pwa to your agent. Access on your computer or phone (via vpn/tailscale). See demo at https://hotbutter.ai - Embeds local relay server and PWA for secure, self-hosted voice chat. - Terminal provides a 6-digit pairing code and one-click URL to start sessions. - Real-time speech-to-text (STT) in browser; messages relayed to agent and back. - Agent responses delivered by browser text-to-speech (TTS). - Supports configuration of server port and agent display name via CLI.
Metadata
Slug hotbutter
Version 1.0.6
License
All-time Installs 0
Active Installs 0
Total Versions 6
Frequently Asked Questions

What is hotbutter voice chat?

Enables local voice chat by embedding Hotbutter relay server and PWA, providing speech-to-text and text-to-speech via a secure, self-hosted connection. It is an AI Agent Skill for Claude Code / OpenClaw, with 490 downloads so far.

How do I install hotbutter voice chat?

Run "/install hotbutter" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is hotbutter voice chat free?

Yes, hotbutter voice chat is completely free (open-source). You can download, install and use it at no cost.

Which platforms does hotbutter voice chat support?

hotbutter voice chat is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created hotbutter voice chat?

It is built and maintained by Michael Stajer (@michael-stajer); the current version is v1.0.6.

💬 Comments