WakeHermesClaw for Android 🦞

A native Android voice client for OpenClaw and Hermes Agent.

WakeHermesClaw is the successor to OpenClaw Assistant. It keeps every OpenClaw feature you already have — wake word, Voice Overlay, Gateway

• HTTP backends, Wear OS, on-device node capabilities, continuous conversation — and adds first-class support for Hermes Agent as a peer backend that can be configured side-by-side with OpenClaw.

Supported backends (configure any combination, pick one as primary):

• OpenClaw Gateway (WebSocket, pairing, QR, TLS)
• OpenClaw HTTP (OpenAI-compatible chat)
• Hermes API Server (/v1/chat/completions streaming + /v1/runs Runs API)

Mobile Bridge (optional, off by default). WakeHermesClaw can expose a bearer-token-protected local HTTP service that lets Hermes reach a curated set of Android capabilities. See docs/hermes-mobile-bridge.md and integrations/hermes-mobile-bridge/.

Migration. Existing OpenClaw Assistant installs upgrade in place (applicationId unchanged); the first launch silently migrates the previous Gateway / HTTP settings into the new multi-backend repository and marks the existing setup as Primary. Wake word, HotwordService, Voice Overlay, Wear OS, and node capabilities continue to target the Primary backend.

Hermes setup in 30 seconds. Run hermes gateway with the API server enabled (default port 8642), bind it to your LAN/VPN, generate an API key, and add it in Settings → Backends → Add Hermes API Server. WakeHermesClaw accepts both http://host:8642 and http://host:8642/v1. Authentication uses Authorization: Bearer <key>; the default model name is hermes-agent. Connection test calls GET /v1/models and falls back to /health.

WakeHermesClaw shared agent controls. The app includes 6-character bridge pairing codes, deep-link setup payloads, multi-endpoint candidates (LAN + VPN + public) raced in parallel on connect, per-capability TTL grants ("approve · 10 min / 1 hour / until revoked") with a destructive-verb override, /revoke and /grants endpoints, an Accessibility Bridge (tap / swipe / Home / Back / window describe — sideload-only), notifications.active.list backed by the existing NotificationListenerService, and a build-flag (IS_SIDELOAD) that gates Accessibility + SMS for a Play-track build.

Wear OS uses its own per-watch backend URL + auth token. Point the watch at the same Hermes (http://host:8642, Bearer = your API key) or OpenClaw endpoint you use on the phone. The phone's Primary backend is not automatically synced to the watch yet.

Build: ./gradlew testDebugUnitTest assembleDebug.

OpenClaw Assistant 🦞 (legacy README continues below)

日本語版はこちら | English below

Demo

Screenshots

English

Your AI Assistant in Your Pocket - A self-hosted Android voice assistant for OpenClaw with offline wake word detection, real device actions, and a UI built for everyday use.

Why OpenClaw Assistant

• Feels native on Android - Long press Home, use wake word detection, stream replies, and keep talking in continuous conversation mode
• Actually controls your device - Notifications, camera, contacts, calendar, apps, clipboard, WiFi, screen sharing, and more
• Self-hosted, not locked to one cloud app - Connects to your OpenClaw gateway over WebSocket or HTTP
• Works beyond the phone - Includes Wear OS support and a Canvas tab for richer AI-driven UI

🚀 5-Minute Quick Start

1. Install the latest APK from Releases
2. On your server, run openclaw qr
3. Open the app and tap Scan QR Code
4. Approve the device on your server if pairing is requested
5. Long press Home or say your wake word

Need the server first? Start with OpenClaw.

🔒 Privacy & Security at a Glance

• Offline wake word detection runs locally on-device with Vosk
• Sensitive settings like server URLs and tokens are stored with AES256-GCM encryption
• Powerful capabilities are opt-in and guarded by Android permissions or explicit user actions

✨ Features

Voice & Speech

• 🎤 Customizable Wake Word - Choose from "OpenClaw", "Hey Assistant", "Jarvis", "Computer", or set your own custom phrase
• 📴 Offline Wake Word Detection - Always-on local processing powered by Vosk, no internet required
• 🗣️ Speech Recognition - Real-time speech-to-text with partial results display and configurable silence timeout
• 🔊 Text-to-Speech - Automatic voice output with adjustable speech speed, multi-engine support, and smart text chunking for long responses. Supports Android native, ElevenLabs, OpenAI, and VOICEVOX (full build) TTS providers
• 🔄 Continuous Conversation Mode - Auto-resumes listening after AI response for natural back-and-forth dialogue
• 🏠 System Assistant Integration - Long press Home button to activate via Android VoiceInteractionService
• 🔃 Wake Word Sync - Download wake words configured on the gateway server to your device
• ⌚ Wear OS Support - Use OpenClaw Assistant directly from your smartwatch

Chat & AI

• 💬 In-App Chat Interface - Full-featured chat UI with text and voice input, markdown rendering, and message timestamps
• 🤖 Agent Selection - Choose from multiple AI agents fetched dynamically from the gateway
• 📡 Real-time Streaming - See AI responses as they are generated via WebSocket gateway
• 💾 Chat History - Local message persistence with session management (create, switch, delete conversations)
• 🔔 Thinking Sound - Optional audio cue while waiting for AI response
• 🪟 Dual Chat Modes - Gateway Chat (via Node-Gateway connection) or HTTP Chat (direct HTTP endpoint)
• 📎 File Attachments - Attach images directly in the chat interface

Gateway & Connectivity

• 🌐 WebSocket Gateway - Persistent connection with auto-reconnect (exponential backoff), ping keep-alive, and RPC protocol
• 🔍 Auto-Discovery - Automatically find OpenClaw gateways on your local network via mDNS/Bonjour
• 🔌 Manual Connection - Specify host, port, and token for direct connection
• 🔒 TLS Support - Encrypted connections with SHA-256 fingerprint verification dialog for first-time trust
• 📋 Agent Discovery - Dynamically fetch available agents from the gateway
• 🔗 Device Pairing - Server-side device approval with Ed25519 cryptographic identity
• ✅ Connection Testing - Built-in connection test with live feedback in settings

Node Capabilities

• 📷 Camera - AI can capture photos and video clips via the device camera
• 📍 Location - Share your location (Off / Coarse / Precise) with the AI, including background tracking
• 📲 SMS - Allow the AI to send and read text messages with your permission
• 🖥️ Screen Recording - Let the AI see your screen when you explicitly ask it to
• 🔔 Notifications - AI can read and interact with your active notifications
• 🏃 Motion & Activity - Step counting and physical activity detection (walking, running, etc.)
• 📶 WiFi - AI can check WiFi status and list available networks
• 👥 Contacts - AI can search, add, update, and delete contacts
• 📅 Calendar - AI can query, add, update, and delete calendar events
• 📋 Clipboard - AI can read from and write to the clipboard
• 📱 Apps - AI can list installed apps and launch applications
• 🖼️ Photos - AI can retrieve recent photos from your device gallery
• 🔧 System Control - AI can adjust volume, screen brightness, and send notifications

System & Security

• 🔒 **Encrypted Settings