How to Setup OpenClaw for FREE - Local AI Agent Using Ollama (Full Tutorial)
Tired of AI assistants that just talk? OpenClaw transforms WhatsApp and Telegram into powerful action-taking agents that manage your files, emails, and computer tasks - all running privately on your machine. This step-by-step guide shows how to bypass cloud services and API limits using Ollama's local LLMs.
What Makes OpenClaw Different?
Most AI assistants today are glorified chatbots - they answer questions but can't actually do anything. OpenClaw (also called Clawdbot or Moltbot) changes this by turning your messaging apps into powerful action interfaces. Imagine clearing your inbox, finding files, or running computer commands - all through natural conversation in WhatsApp or Telegram.
Unlike cloud-based services, OpenClaw runs locally on your machine. This means no API limits, no subscription fees, and complete privacy for your data. When combined with Ollama's local LLMs, you get an AI assistant that's both powerful and private.
Key differentiator: OpenClaw isn't just an information source - it's an execution engine. Where ChatGPT tells you how to create a file, OpenClaw actually creates it when you ask through WhatsApp.
Step 1: Installing OpenClaw
The installation process is straightforward but requires careful attention to terminal commands. At the 2:15 mark in the video, you'll see the exact installation steps:
- Search for "open claw" on Google and navigate to the official website
- Scroll down to find the installation command
- Copy the command to your clipboard
- Open terminal and paste the installation command
The installation typically takes 10-15 minutes depending on your internet speed. Once complete, you'll see the setup screen where you'll need to:
- Accept the terms and conditions
- Select "Quick Start" for initial configuration
- Choose your preferred model (we'll use Google Gemini initially)
Pro Tip: Keep a notepad open to save API keys and tokens during setup. You'll need these for multiple steps.
Step 2: Configuring Google Gemini
While we'll eventually switch to local Ollama models, starting with Gemini provides a reliable baseline. Here's how to get your API key:
- Search for "Gemini API key" on Google
- Sign in to the Google AI Studio website
- Navigate to "Get API Key" in the bottom left
- Click "Create API Key" and name it (e.g., "OpenClaw")
- Copy the generated key to your clipboard
Back in your terminal during OpenClaw setup:
- Paste the Gemini API key when prompted
- Select "Gemini 2.5" as your default model
- Proceed to channel configuration
Important: The free Gemini tier has usage limits. We'll bypass these later by switching to Ollama, but the initial setup ensures everything works before making it fully local.
Step 3: Creating Your Telegram Bot
Telegram integration turns your messaging app into an AI command center. The setup involves:
- Searching for "BotFather" in Telegram
- Starting a chat and selecting "/newbot"
- Naming your bot (must end with "_bot")
- Receiving and saving your unique bot token
In your OpenClaw terminal setup:
- Paste the Telegram bot token when prompted
- Skip skills and webhooks for now (we'll add these later)
- Wait for gateway installation to complete
Once installed, you'll need to:
- Open your Telegram bot and send "/start"
- Approve the pairing request in your terminal
- Begin chatting with your new AI assistant
Security Note: The initial pairing approval is crucial - it prevents unauthorized access to your machine through the bot.
Testing File Commands
At the 7:30 timestamp in the video, you'll see the first real test of OpenClaw's capabilities. Try these commands in your Telegram bot:
-
Create a file named welcome.txt with content "Hello from OpenClaw" -
List all PDFs downloaded last week -
Show me the file welcome.txt
Initially, you may encounter permission errors. This is a security feature - OpenClaw requires explicit approval before accessing personal files. To grant permissions:
- Prefix commands with
command(e.g.,command ls Downloads) - Approve the action in your terminal when prompted
- Subsequent commands to the same location won't require re-approval
Real-world use case: At 8:45 in the video, the bot successfully creates an index.html file with custom content - proving OpenClaw can handle web development tasks through chat.
Step 4: WhatsApp Integration
Adding WhatsApp follows a similar but distinct process:
- Run
openclaw onboardin terminal - Accept agreements and select "Quick Start"
- Choose "Update values" for config handling
- Skip model configuration (we're keeping Gemini for now)
- Select WhatsApp from the channel list
The critical WhatsApp-specific steps:
- Scan the QR code with your phone (WhatsApp → Settings → Linked Devices)
- Use a secondary phone number if possible
- Input your number when prompted in terminal
- Skip skills configuration for now
- Restart the gateway
At the 11:20 mark, you'll see the WhatsApp integration maintaining context from Telegram - proof that OpenClaw creates a unified AI assistant across platforms.
Business application: This allows employees to access company files or data through WhatsApp without installing special software - just chat with your AI assistant.
Step 5: Local AI with Ollama
When you hit API limits with Gemini (shown at 12:40 in the video), Ollama provides the solution:
- Search for "ollama download" and get the appropriate version
- Install and launch the Ollama application
- Select a model (Llama 2, Mistral, etc.) from the interface
- Initiate a chat to trigger model download
Key advantages of Ollama:
- Runs completely locally - no internet required after setup
- No API limits or costs
- Multiple model options for different needs
- Persistent memory across sessions
Performance tip: Smaller models like Mistral 7B work well for most assistant tasks, while larger models are better for complex reasoning. The video shows model selection at 13:15.
Critical Configuration Updates
The most technical but essential step comes at 14:30 - modifying OpenClaw's config.json to work with Ollama:
- Navigate to
~/.openclawdirectory - Open
config.jsonin a text editor - Add the Ollama provider snippet before the agents section:
{ "providers": [ { "name": "ollama", "base_url": "http://localhost:11434", "api_key": "anything", "models": ["llama2"] } ] } - Save the file and rerun
openclaw onboard - Select Ollama as your provider when prompted
After restarting the gateway, test with:
-
What models do you have access to? -
Find all spreadsheets modified yesterday -
Create a meeting note for tomorrow at 3pm
Success indicator: At 15:50 in the video, the bot successfully lists PDF files using Ollama - now your AI assistant is fully local and private.
Watch the Full Tutorial
For visual learners, the video tutorial shows real-time troubleshooting (like the permission issues at 9:10) and advanced configurations not covered in this article. See how the WhatsApp integration maintains context from Telegram at 11:20, and watch the Ollama model selection process at 13:15.
Key Takeaways
OpenClaw transforms everyday messaging apps into powerful AI assistants that actually do work rather than just talk about it. By combining it with Ollama's local LLMs, you create a private, capable assistant that:
- Operates without cloud dependencies or API limits
- Maintains context across WhatsApp and Telegram
- Executes real computer tasks through natural language
- Protects your data by keeping everything local
In summary: In about 30 minutes, you can deploy a private AI agent that handles files, emails, and system tasks through WhatsApp/Telegram - no subscriptions, no data leaks, just productive automation.
Frequently Asked Questions
Common questions about OpenClaw and local AI agents
OpenClaw runs locally on your machine rather than in the cloud, giving you complete privacy and control. Unlike SaaS AI tools, it integrates directly with messaging apps like WhatsApp/Telegram to perform actual tasks like file management, email handling, and system commands without requiring you to use a separate dashboard.
The assistant maintains context across conversations and platforms, remembering your preferences and past interactions. This creates a truly personalized experience that improves with use.
- Action-oriented - Doesn't just answer questions, executes tasks
- Private by design - No data leaves your machine
- Messaging app integration - Works where you already communicate
No advanced technical skills are required. The setup involves copying a few commands into your terminal and following the guided configuration. The most technical part is obtaining API keys, which involves simple web form submissions.
The entire process takes about 30 minutes for first-time users. The video tutorial shows each step visually, and the configuration menus provide clear options rather than requiring manual coding.
- Terminal commands are provided - just copy/paste
- Configuration is menu-driven with clear options
- Error messages include suggested fixes
Ollama provides free, locally-run large language models that avoid API rate limits and costs associated with cloud providers like Google Gemini. When configured with OpenClaw, it creates a completely private AI assistant that doesn't send your data to third-party servers.
Popular Ollama models like Llama 2 and Mistral work well for this setup. They offer a good balance of performance and resource usage, with smaller models running well even on modest hardware.
- No usage limits - Unlike cloud APIs with quotas
- Works offline - Once models are downloaded
- Multiple model options - Choose based on your needs
The assistant can manage files (create/edit/retrieve), execute system commands, schedule calendar events, draft/send emails, check flights, and automate repetitive computer tasks. All actions are triggered through natural conversation in your messaging apps.
Advanced capabilities include data analysis (working with spreadsheets), web research (with proper configuration), and even controlling smart home devices when integrated with appropriate platforms.
- File management - Create, edit, organize, retrieve
- System control - Run commands, manage processes
- Communication - Draft emails, summarize messages
OpenClaw includes multiple security layers. It requires explicit permission for file access, validates all commands through confirmation prompts, and operates within a sandboxed environment. For sensitive operations, you can restrict folder access in the configuration file.
The local setup means your data never leaves your machine. Communication with WhatsApp/Telegram uses the same encryption those platforms normally employ, with no additional data collection.
- Explicit permissions - Per-folder access control
- Command validation - Confirms potentially risky actions
- No data sharing - Everything stays on your devices
Yes, OpenClaw supports all major operating systems. The installation process differs slightly between platforms (Homebrew for Mac, apt for Linux, etc.), but the core functionality remains identical. The tutorial covers setup for all environments with platform-specific commands where needed.
Performance may vary based on hardware capabilities, especially when running local LLMs through Ollama. More powerful systems can handle larger models and more concurrent tasks.
- Windows: PowerShell installation
- Mac: Homebrew package manager
- Linux: apt or other native package tools
While ChatGPT/Gemini are conversational AIs, OpenClaw is an action-oriented agent. Instead of just answering questions, it performs actual computer tasks through natural language commands. The local Ollama integration provides similar conversational abilities without the cloud dependency, API costs, or privacy concerns of commercial services.
For businesses, the key advantage is workflow integration. OpenClaw can become part of your daily operations rather than just a reference tool, automating repetitive tasks across your systems.
- Action vs information - Does tasks rather than explains them
- Private operation - No data shared with third parties
- No usage costs - After initial setup
GrowwStacks specializes in deploying private AI automation solutions for businesses. We can configure OpenClaw with custom skills for your workflows, integrate it with your existing tools (CRM, email, docs), and deploy secure team instances.
Our implementation service includes security hardening, performance optimization, and training for your team. We handle the technical setup so you can focus on using the assistant to improve productivity.
- Custom workflows - Tailored to your business processes
- Team deployment - Secure multi-user configuration
- Ongoing support - Updates and troubleshooting
Ready to Deploy Your Private AI Assistant?
Manual setup works for individuals, but businesses need reliable, secure implementations. GrowwStacks delivers turnkey OpenClaw deployments with custom skills, team access controls, and enterprise-grade security.