🤖 ClawBot Plus — AI-Powered Computer Control Agent

Your PC. Your AI. Full Control.

ClawBot Plus is the ultimate personal AI assistant and open-source autonomous computer control agent for your Windows PC. Whether you need a powerful OpenClaw alternative, a browser automation tool using Playwright, or a Voice AI agent that talks back to you via Edge TTS, ClawBot Plus does it all. It can autonomously open apps, write code, run Python scripts, generate PowerPoint presentations, scrape the web, control system settings, and manage a multi-agent swarm — all from a single CLI or a remote Telegram bot.

If you are looking for an AI that feels local, fast, persistent, and acts as your robotic pair programmer and RPA (Robotic Process Automation) engine — this is it.

Install (recommended)

Runtime: Python 3.11+

pip install clawbot-plus
clawbot

That's it. ClawBot will auto-detect your system, scan your apps, and set up its workspace on first run.

Quick Start

# Start ClawBot
clawbot

# Choose your AI provider (Gemini, OpenAI, Groq, Ollama, GitHub Models, NVIDIA NIM, etc.)
# Choose your model
# Start giving commands!

❯ open chrome and go to youtube
❯ make a portfolio website
❯ what's the weather today
❯ /voice   # Start voice mode
❯ switch   # Change AI model
❯ quit     # Exit

From Source (development)

git clone https://github.com/aayushsoam/clawbot-plus.git
cd clawbot-plus
pip install -e .
clawbot

Supported AI Providers

Provider	Models	Auth
🔷 Google Gemini	gemini-3.1-pro, gemini-2.5-flash, gemini-2.0-flash, etc.	GEMINI_API_KEY
🔵 OpenAI	gpt-4o, gpt-4.1-mini, o3-mini, o4-mini	OPENAI_API_KEY
🟠 Anthropic	claude-3-7-sonnet, claude-3-5-sonnet, claude-3-opus	ANTHROPIC_API_KEY
⚡ Groq	llama-3.3-70b-versatile, gemma2-9b-it	GROQ_API_KEY
🐋 DeepSeek	deepseek-chat, deepseek-reasoner	DEEPSEEK_API_KEY
🟩 NVIDIA NIM	nemotron-120b, llama-3.3-70b, deepseek-r1, phi-4	NVIDIA_API_KEY
🌙 Kimi (Moonshot)	moonshot-v1-8k/32k/128k	MOONSHOT_API_KEY
🐙 GitHub Models	gpt-4o, gpt-4.1, Codestral-2501, Llama-3.3-70B	GITHUB_TOKEN
🦙 Ollama (Local)	qwen3-vl, llama3.3, deepseek-r1, mistral, phi3	Free (local)
☁️ Ollama (Cloud)	Same as local, custom host	OLLAMA_BASE_URL

Highlights

• 🖥️ Full Computer Control — open apps, click, type, scroll, screenshot, drag & drop. ClawBot sees your screen and acts.
• 💻 Senior Developer Coding — writes production-quality code with separate HTML/CSS/JS files, verifies via shell, and shows output.
• 🔊 Voice Mode — /voice to talk to ClawBot using your microphone. It speaks back using Edge TTS.
• 📱 Telegram Bot — connect your Telegram bot and control your PC from your phone.
• 🧠 Persistent Memory — remembers your name, preferences, past sessions, and daily logs across restarts.
• 🛡️ Anti-Loop Engine — detects when the AI is stuck and forces a different approach or graceful exit.
• 🔍 Web Search — search the web and fetch pages without opening a browser.
• 📊 PPT Generation — generate_ppt action creates professional PowerPoint presentations.
• 🎬 Video Generation — local Python or cloud API video creation via Koda-Stack.
• 🤖 Multi-Agent Swarm — spawn worker agents for complex projects (ClawTeam).
• 🔒 Security Layer — dangerous commands are blocked. All shell commands are audited.
• 📦 Skills System — extensible SKILL.md-based plugin architecture for CCTV, CapCut, LaTeX, Office docs, and more.

How It Works

                            ╭──────────────────────────────╮
                            │  You (CLI / Telegram / Voice)│
                            ╰──────────────┬───────────────╯
                                           │ (Commands, Voice, Images)
                                           ▼
         ╭───────────────────────────────────────────────────────────────────╮
         │ 🤖 ClawBot Plus (Agentic Core)                                    │
         │   • Brain (LLMs)      • RAG Memory       • Anti-Loop Engine       │
         │   • Skill Plugins     • Error Healing    • Multi-Agent Team       │
         ╰───────────────────────────────┬───────────────────────────────────╯
                                         │
             ╭───────────────────────────┼───────────────────────────╮
             │                           │                           │
             ▼                           ▼                           ▼
 ╭───────────────────────╮   ╭───────────────────────╮   ╭───────────────────────╮
 │ 🖥️ System Control     │   │ 💻 Code Engine        │   │ 🌐 Browser Automation │
 │ --------------------- │   │ --------------------- │   │ --------------------- │
 │ • OS APIs / PyAutoGUI │   │ • read/write_file     │   │ • browser-use Agent   │
 │ • UI Automation       │   │ • Shell (Subprocess)  │   │ • Playwright / DOM    │
 │ • Window Management   │   │ • Syntax Validation   │   │ • Scrape & Extract    │
 ╰───────────────────────╯   ╰───────────────────────╯   ╰───────────────────────╯

Key Subsystems

• Computer Agent (clawbot/computer/agent.py) — the brain. Receives tasks, plans actions, executes them, and learns.
• CLI (clawbot/cli.py) — the face. Interactive prompt with model selection, mode switching, and live task execution.
• Telegram Bot (clawbot/computer/telegram_bot.py) — remote control from your phone.
• Voice Engine (clawbot/computer/voice.py) — speech-to-text + text-to-speech loop.
• Skills (skills/) — modular capabilities loaded on-demand based on task keywords.
• Memory (~/.clawbot/) — sessions, daily logs, preferences, and persona files.

Workspace & Persona

• Workspace root: ~/.clawbot/workspace/
• Persona files: IDENTITY.md, SOUL.md, USER.md, AGENTS.md, BOOTSTRAP.md
• Skills: skills/<skill_name>/SKILL.md
• Sessions: ~/.clawbot/sessions/ (one .md per task)
• Daily memory: ~/.clawbot/memory/ (one .md per day)

Configuration

Minimal ~/.clawbot/clawbot.json:

{
  "default_provider": "🔷 Google Gemini (Free)",
  "default_model": "gemini-2.5-flash"
}

CLI Commands

Command	Action
clawbot	Start the interactive CLI
clawbot --version	Show version
clawbot devices	Manage Claw3D device connections
clawbot tunnel	Setup remote access tunnel
/voice	Start voice control mode
/video	Switch to video generation mode
/code	Route tasks to OpenCode
/normal	Back to standard agent mode
switch	Change AI provider/model
quit	Exit ClawBot

Chat Commands (Telegram)

Send these from Telegram when connected:

• Any text → executes as a task on your PC
• Images → analyzed by vision models
• Voice messages → transcribed and executed

Tech Stack

• Python 3.11+ — core runtime
• browser-use — LLM classes and browser automation
• PyAutoGUI — screen control, clicks, typing
• pywin32 — Windows native API (UIAutomation)
• Rich — beautiful terminal UI
• InquirerPy — interactive prompts
• edge-tts — text-to-speech
• SpeechRecognition — voice input
• python-telegram-bot — Telegram integration
• Pillow — image processing for vision
• httpx — async HTTP client

Project Structure

clawbot/
  __init__.py
  cli.py              # Main CLI entry point
  computer/
    agent.py           # Core AI agent loop
    telegram_bot.py    # Telegram integration
    voice.py           # Voice mode
    actions.py         # System actions
    system_control.py  # OS-level controls
    window_manager.py  # Window focus/snap
    security.py        # Command auditing
    gateway.py         # Claw3D WebSocket gateway
    memory_search.py   # RAG-based memory search
    self_healing.py    # Auto-fix failed code
    workflows.py       # Macro recording/replay
    orchestrator.py    # Multi-agent coordination
  clawteam/            # Multi-agent swarm system
skills/                # Extensible skill plugins
workspace/             # Persona & identity files
agents/                # AI reviewer/planner agents

Star History

Built By

ClawBot Plus was built by Aayush Soam with significant contributions from Claude (Anthropic) — pair-programmed from architecture to deployment.

• 👤 Aayush Soam — Creator, Vision & Product
• 🤖 Claude (Anthropic) — Co-Developer, Architecture & Agent Engineering

License

MIT