📺 Watch the full tutorial on YouTube

OpenClaw 2.0 Dropped! Setup Your Own Telegram AI Agent Now

OpenClaw 2.0 is a local-first open-source AI agent workspace that connects language models (local Ollama or Google Gemini) to real-world software tools, files, browsers, and Telegram messaging channels with 100% data privacy.

Step 1: Input	Step 2: AI Action	Step 3: Result
User prompt or automated trigger via Telegram bot or CLI TUI	OpenClaw 2.0 routes task to selected model (Ollama or Gemini)	Safe tool execution, credential masking, and durable session state output

---

⚡ Quick 1-Line Installation

Install OpenClaw 2.0 on Windows instantly using PowerShell:

iwr -useb https://openclaw.ai/install.ps1 | iex

Alternative via Node/npm:

npm install -g openclaw@latest

---

🚀 Key Features

• 📱 Mobile Telegram Channel Connectivity: Interact with your AI agent on Telegram with instant response notifications and human approval prompts.
• 🧠 Dual Provider Support: Seamlessly switch between local Ollama models (llama3.2:3b, qwen3:4b) and Google Gemini (gemini-3.5-flash-lite).
• 🔑 CLI Secret Security (No .env required): API keys and tokens are stored securely in encrypted OpenClaw auth profiles via CLI commands.
• 📂 Remote PC File Operations: Read, write, search, and edit files on your PC remotely through Telegram.
• 🛡️ Masked Credential Protection: Replaces sensitive API keys with temporary secrets prior to prompt processing, ensuring zero raw secret leakage.
• ⚙️ Exact-Operation Approvals: Inspect, grant, or revoke permission per single task action instead of granting permanent blanket access.

---

💻 Tech Stack

• Agent Runtime: OpenClaw v2026.8.1
• Supported Providers: Local Ollama (/api/chat) & Google Gemini API
• Messaging Channel: Telegram Bot API
• CLI Commands: openclaw onboard, openclaw models, openclaw channels, openclaw gateway

---

🛠️ Step-by-Step Setup Guide

1. Run Guided Onboarding Wizard

Initialize OpenClaw workspace, auth profiles, and core settings:

openclaw onboard

• When to run: On initial system installation to set up default workspaces, gateway auth tokens, skills, and channels interactively.

---

2. Configure Model Provider (Choose Option A or B)

OpenClaw supports both Local Ollama models and Google Gemini:

Option A: Local Ollama Models (100% Offline & Free)

1. Install & start Ollama from ollama.com and pull models:

   ollama pull llama3.2:3b
   ollama pull qwen3:4b

2. Set Ollama as your default active model:

   openclaw models set ollama/llama3.2:3b

Option B: Google Gemini 3.5 Flash Lite (Cloud Performance)

1. Securely pass your Gemini API Key directly via OpenClaw CLI:

   # Interactive prompt:
   openclaw models auth paste-api-key --provider google
   
   # Or 1-line command:
   "YOUR_GEMINI_API_KEY" | openclaw models auth paste-api-key --provider google

2. Set Gemini 3.5 Flash Lite as your default active model:

   openclaw models set google/gemini-3.5-flash-lite

---

3. Switch Between Gemini & Ollama On-The-Fly

You can dynamically switch primary models anytime via CLI or Telegram chat commands:

• Switch to Google Gemini 3.5 Flash Lite:

  openclaw models set google/gemini-3.5-flash-lite

  (Or send /model google/gemini-3.5-flash-lite in Telegram or CLI TUI)

• Switch to Local Ollama Model:

  openclaw models set ollama/llama3.2:3b

  (Or send /model ollama/llama3.2:3b in Telegram or CLI TUI)

---

4. Link Telegram Channel & Enable Full Tools

1. Add Telegram Bot Token:

   openclaw channels add --channel telegram --token 8659491511:YOUR_TELEGRAM_BOT_TOKEN

2. Enable Full File & Remote PC Tools (Required for reading/saving files):

   openclaw config set tools.profile full

   • Why this is required: Ensures OpenClaw has permissions to read, write, search, and edit files on your PC when requested via Telegram.

3. Set Project Workspace Directory:

   openclaw config set agents.defaults.workspace "d:\Ray Codes\AG Projects\OpenClaw 2.0"

   • Why this is required: Directs OpenClaw to save all generated scripts and read workspace files directly inside your active project folder.

4. Allow Telegram Direct Messages:

   openclaw config set channels.telegram.dmPolicy open

---

5. Start the OpenClaw Gateway Listener

To start listening to incoming Telegram messages and executing agent tasks in real-time:

openclaw gateway run

• When to run: Whenever you want OpenClaw actively listening and responding to prompts on Telegram.
• What it does: Launches the WebSocket gateway, establishes Telegram polling/webhooks, and executes tool actions.

---

6. Run in Background as a Service (Optional)

To keep OpenClaw running continuously in the background 24/7:

openclaw daemon install
openclaw daemon start

---

7. Check Connection & Model Status

Verify gateway, channel health, and active model configuration:

openclaw status
openclaw channels status
openclaw models status

---

📋 OpenClaw CLI Quick Command Reference

Command	Action
openclaw onboard	Starts interactive guided setup wizard
openclaw gateway run	Starts OpenClaw Gateway in foreground listening for Telegram
openclaw channels add --channel telegram --token <TOKEN>	Registers Telegram bot channel non-interactively
openclaw models auth paste-api-key --provider google	Prompts interactively to paste Google Gemini API key
"YOUR_API_KEY" | openclaw models auth paste-api-key --provider google	Sets Google Gemini API key non-interactively in 1 line
openclaw models set google/gemini-3.5-flash-lite	Switches active model to Gemini 3.5 Flash Lite
openclaw models set ollama/llama3.2:3b	Switches active model to local Ollama
openclaw config set tools.profile full	Enables full file reading, writing, and system tool permissions for Telegram
openclaw config set agents.defaults.workspace <PATH>	Sets active project folder where Telegram files are read and created
openclaw config set channels.telegram.dmPolicy open	Configures Telegram to allow direct message execution
openclaw daemon start	Runs OpenClaw 24/7 as a background Windows Service

| openclaw status | Shows Gateway connectivity, channels, and active model health |

---

📱 How to Interact with OpenClaw on Telegram

Once OpenClaw Gateway is running (openclaw gateway run), you can send prompts, run model commands, and remotely manage PC files directly from Telegram:

📁 Remote PC File Operations via Telegram

Example 1: Reading a File on Your PC Remotely

• You send on Telegram: "Read the notes.txt file from my project workspace and summarize key points."
• OpenClaw Action: Opens notes.txt on your PC using local tools and replies with a summary on Telegram.

Example 2: Creating or Editing a File on Your PC

• You send on Telegram: "Write a Python script for system health monitoring and save it as system_health.py in my workspace."
• OpenClaw Action: Generates the script using active LLM, writes system_health.py directly to your PC workspace, and confirms on Telegram.

---

🤖 Model Commands & Multi-Modal Features

1. On-the-Fly Model Switching (/model)

• Send /model to view all active Ollama & Gemini models.
• Send /model google/gemini-3.5-flash-lite to switch to Gemini cloud reasoning.
• Send /model ollama/llama3.2:3b to switch back to local offline model.

2. Mobile Operation Approvals & Security

• When the agent attempts to execute sensitive commands or edit core files on your PC, OpenClaw sends an interactive Telegram confirmation card:
  • Press [Approve Operation] to allow execution on your PC.
  • Press [Deny] to revoke permission.

---

📂 File Architecture

OpenClaw 2.0/
├── README.md          # Complete project documentation and Telegram guide
├── notes.txt          # Workspace notes
├── system_health.py   # System health monitoring script example
└── assets/
    ├── openclaw_2.0_overview.html   # Visual HTML presentation asset
    └── openclaw_2.0_overview.md     # Matching Markdown companion report

---

🏷️ Keywords & Tags

OpenClaw 2.0 Multiplayer AI Coding Agent Local AI Workstation Ollama Integration Google Gemini Telegram Bot Gateway Autonomous AI Agent Open Source AI Local LLM Orchestration AI Dev Tools