Waveform Intelligence Engine
Skeeter is a personal AI assistant platform built around Claude. It runs as a CLI daemon and connects to multiple channels — Telegram, web chat, and terminal — routing messages through a shared agent runtime with persistent memory, scheduled tasks, voice I/O, and an extensible tool system.
skeeter/
├── src/
│ ├── agent/ # Core agent runtime, streaming, context, prompt
│ │ ├── providers/ # Model provider abstraction
│ │ └── tools/ # All agent tools (see below)
│ ├── channels/ # Input/output channels
│ │ ├── cli/ # Terminal chat
│ │ ├── telegram/ # Telegram bot
│ │ └── webchat/ # Browser-based chat UI
│ ├── brain/ # Persistent knowledge store
│ ├── memory/ # Session and long-term memory (Moonshine)
│ ├── planner/ # Goal decomposition and task scheduling
│ ├── cron/ # Recurring task scheduler
│ ├── calendar/ # Calendar awareness
│ ├── identity/ # Agent persona and identity config
│ ├── session/ # Conversation session management
│ ├── stt/ # Speech-to-text (Moonshine, local)
│ ├── tts/ # Text-to-speech (ElevenLabs, VibeVoice)
│ ├── integrations/ # External service integrations
│ │ ├── coolify/ # Self-hosted deployment via Coolify
│ │ └── mission-control/
│ └── cli/ # CLI command definitions
├── landing/ # Static landing page (Grid, Cube, Audio visualiser)
└── tests/ # Vitest test suite
The agent has access to a broad tool surface:
| Category | Tools |
|---|---|
| System | bash, filesystem, workspace |
| Web | web (fetch/scrape), image |
| Memory | memory, brain, genome |
| Planning | planner, cron, calendar |
| Voice | voice-synth, formant-speak, vibevoice, tone |
| Audio | audio-training |
| Visualisation | cube, grid, ultraman |
| Agent | agent, skills |
| Utilities | utility, avatar |
| Language | TypeScript 5, Node.js 22+ |
| AI | Anthropic Claude (via @anthropic-ai/claude-agent-sdk) |
| HTTP | Hono |
| Telegram | grammY |
| Build | tsup |
| Tests | Vitest |
| Package manager | pnpm |
# Install dependencies
pnpm install
# Build
pnpm build
# Run in dev mode
pnpm dev
# Or use the CLI directly after build
node dist/index.js --help
skeeter gateway # Start the channel gateway (all channels)
skeeter agent # Start agent in terminal
skeeter config # Manage configuration
skeeter devices # Manage audio devices
skeeter channels # Manage active channels
skeeter cron # Manage scheduled tasks
Create a .env file at the project root:
ANTHROPIC_API_KEY=your_key
# Telegram
TELEGRAM_BOT_TOKEN=your_token
# TTS
ELEVENLABS_API_KEY=your_key
# Coolify integration (optional)
COOLIFY_TOKEN=your_token
COOLIFY_URL=https://your-coolify-instance
pnpm test
1 commits
HTML
59.3%
TypeScript
35.2%
JavaScript
5.6%
Waveform Intelligence Engine
Skeeter is a personal AI assistant platform built around Claude. It runs as a CLI daemon and connects to multiple channels — Telegram, web chat, and terminal — routing messages through a shared agent runtime with persistent memory, scheduled tasks, voice I/O, and an extensible tool system.
skeeter/
├── src/
│ ├── agent/ # Core agent runtime, streaming, context, prompt
│ │ ├── providers/ # Model provider abstraction
│ │ └── tools/ # All agent tools (see below)
│ ├── channels/ # Input/output channels
│ │ ├── cli/ # Terminal chat
│ │ ├── telegram/ # Telegram bot
│ │ └── webchat/ # Browser-based chat UI
│ ├── brain/ # Persistent knowledge store
│ ├── memory/ # Session and long-term memory (Moonshine)
│ ├── planner/ # Goal decomposition and task scheduling
│ ├── cron/ # Recurring task scheduler
│ ├── calendar/ # Calendar awareness
│ ├── identity/ # Agent persona and identity config
│ ├── session/ # Conversation session management
│ ├── stt/ # Speech-to-text (Moonshine, local)
│ ├── tts/ # Text-to-speech (ElevenLabs, VibeVoice)
│ ├── integrations/ # External service integrations
│ │ ├── coolify/ # Self-hosted deployment via Coolify
│ │ └── mission-control/
│ └── cli/ # CLI command definitions
├── landing/ # Static landing page (Grid, Cube, Audio visualiser)
└── tests/ # Vitest test suite
The agent has access to a broad tool surface:
| Category | Tools |
|---|---|
| System | bash, filesystem, workspace |
| Web | web (fetch/scrape), image |
| Memory | memory, brain, genome |
| Planning | planner, cron, calendar |
| Voice | voice-synth, formant-speak, vibevoice, tone |
| Audio | audio-training |
| Visualisation | cube, grid, ultraman |
| Agent | agent, skills |
| Utilities | utility, avatar |
| Language | TypeScript 5, Node.js 22+ |
| AI | Anthropic Claude (via @anthropic-ai/claude-agent-sdk) |
| HTTP | Hono |
| Telegram | grammY |
| Build | tsup |
| Tests | Vitest |
| Package manager | pnpm |
# Install dependencies
pnpm install
# Build
pnpm build
# Run in dev mode
pnpm dev
# Or use the CLI directly after build
node dist/index.js --help
skeeter gateway # Start the channel gateway (all channels)
skeeter agent # Start agent in terminal
skeeter config # Manage configuration
skeeter devices # Manage audio devices
skeeter channels # Manage active channels
skeeter cron # Manage scheduled tasks
Create a .env file at the project root:
ANTHROPIC_API_KEY=your_key
# Telegram
TELEGRAM_BOT_TOKEN=your_token
# TTS
ELEVENLABS_API_KEY=your_key
# Coolify integration (optional)
COOLIFY_TOKEN=your_token
COOLIFY_URL=https://your-coolify-instance
pnpm test
1 commits
HTML
59.3%
TypeScript
35.2%
JavaScript
5.6%