OpenClaw AI Browser Agents + Lossless Claw + Kimi K2.5 + Nvidia + Ollama

New OpenClaw AI Browser Agent: Automate Anything

Introduction to OpenClaw's New Features

• OpenClaw has released an update allowing AI agents to control the actual Chrome browser, enabling tasks like checking Gmail and browsing LinkedIn without user interaction.

• The presenter will demonstrate how to set up OpenClaw for free in under five minutes, requiring no technical experience.

Setting Up OpenClaw

• The update is based on a new live browser control feature added by Google in the latest version of Chrome, allowing external apps to manage browser sessions.

• Users can run terminal commands to enable remote debugging, which grants full control over their Chrome browser.

Configuration Steps

• Step 1: Enable remote debugging through a provided link. Step 2: Run specific terminal commands to verify functionality.

• After installation, users can choose between two profiles: the "OpenClaw profile" (an isolated session managed by OpenClaw) or the "User profile" (which connects directly to the user's real Chrome session).

Functionality and Permissions

• The User profile allows for direct interaction with personal accounts while maintaining security; permissions are required before any actions are taken by the AI agent.

• Users can automate various tasks such as posting on social media or searching online, although some limitations exist in certain functionalities.

Overview of OpenClaw

• OpenClaw is described as a free open-source agent that operates locally on computers and interfaces with various AI models like GPT and Claude.

• The recent update introduces live browser control capabilities that enhance its functionality beyond previous limitations where it could only operate within a simulated environment.

Conclusion on Browser Modes

• There are three modes available:

• OpenClaw Profile: A separate blank Chrome instance with no personal data.

• User Profile: Direct access to the user's existing Chrome session.

• Chrome Relay: An extension facilitating these interactions.

OpenClaw: New Browser Control Features

Overview of OpenClaw Modes

• Mode Two - User Profile: This mode allows OpenClaw to connect directly to an already open Chrome browser, enabling personalized interactions based on the user's browsing history. It is a new feature introduced in the latest update.

• Chrome Relay Mode: This method uses a Chrome extension to attach to a specific tab manually chosen by the user. It requires more manual input compared to Mode Two, which is automated and more powerful.

Setting Up OpenClaw

• Check Chrome Version: Users must ensure they have Google Chrome version 144 or higher. This can be verified through the help menu in Chrome.

• Enabling Remote Debugging: To set up OpenClaw, users need to type a specific address into their Chrome browser's address bar and toggle a checkbox that enables connection with OpenClaw.

• Terminal Commands: After setting up, users will run commands in their terminal (on Mac, accessed via Command + Space). A popup will appear for permission when OpenClaw attempts to connect.

Using OpenClaw Effectively

• Natural Language Interaction: Users can interact with OpenClaw using simple commands like "go to Google and check AI news." The AI is designed to determine which browser mode to use automatically.

• Caution with Instructions: It's advised not to give open-ended instructions without reviewing them first. Specificity is key for effective use of the tool.

Troubleshooting and Best Practices

• Using Sandbox Profile for Testing: For uncertain tasks or testing purposes, it’s recommended to use the sandbox profile instead of the user profile where sensitive information may be involved.

• Debugging Connection Issues: If there are connection problems, ensure you have updated Chrome and enabled remote debugging. If issues persist on Linux or Windows, refer to troubleshooting documents specific for those systems.

Recap of Key Features

• Summary of Browser Modes: The three modes include:

• OpenClaw: Sandbox environment.

• User Profile: Real browser access (new feature).

• Chrome Relay: Manual tab attachment (less preferred due to bugs).

• Final Setup Steps: To utilize these features effectively:

• Ensure you have Chrome version 144+.

• Enable remote debugging at specified addresses.

• Interact naturally with your AI once connected while being cautious about instructions given.

Additional Resources

• Comprehensive Guides Available: A full step-by-step guide along with prompts for various applications such as lead generation and content creation is available within the video notes from today’s session. Regular updates are provided in this community resource aimed at mastering AI tools effectively.

AI Automation Community and Learning Resources

Overview of AI Automation Learning Opportunities

• The focus is on growing, learning, and scaling with AI automation. A community is available for those interested in these topics.

• Participants can access a structured training program that promises to take them from beginner to expert in AI automation within five weeks, including building an AI agent in under five minutes.

• Daily updates are provided, featuring new guides and video tutorials. Recent content includes a model that can be run for free with OpenClaw.

Coaching and Support Features

• Weekly coaching calls are offered, allowing participants to connect with others locally and gain insights into various aspects of AI automation.

• Questions from the community are addressed during sessions; for example, inquiries about creating sub-agents for fetching news briefings were clarified.

OpenClaw Updates: Memory Management Issues

Challenges with OpenClaw's Memory Functionality

• Users face issues where lengthy conversations lead to loss of context as OpenClaw may erase previous instructions or project history.

• A new plugin called Lossless Claw has been introduced to address these memory problems by providing permanent memory retention.

Introduction of Lossless Claw Plugin

• Lossless Claw allows users' AIs to retain all messages without deletion, enhancing the capability of managing larger projects and searching through chat histories effectively.

• The installation process involves a simple command from GitHub, making it accessible for users looking to improve their OpenClaw experience.

Functionality Enhancements Through Lossless Claw

Technical Improvements Offered by Lossless Claw

• This plugin replaces the existing memory management system with a more efficient method that retains all messages while summarizing older ones instead of deleting them.

• Agents using this system can search through past interactions seamlessly, ensuring no information is lost during conversations.

User Experience Benefits

• With Lossless Claw installed, users will experience an AI that remembers details consistently across interactions, significantly improving usability compared to previous versions.

How to Use Lossless Claw for AI Memory Management

Checking if Lossless Claw is Running

• To verify functionality, enter the chat and ask if "Blossus Claw" is running. It will display details such as version, state, and improvement suggestions.

Benefits of Using Lossless Claw

• This tool enhances memory management for AI agents, making them smarter by addressing issues related to memory handling.

• Unlike traditional systems that truncate old chat history, it retains raw messages and builds summary layers for better continuity in long conversations.

Practical Applications

• Ideal for ongoing projects or research threads where context retention is crucial; it allows revisiting discussions after days without losing coherence.

• Particularly useful during extended coding sessions or when managing extensive conversation histories efficiently.

How Lossless Claw Works

• Conceptually likened to a filing cabinet: instead of discarding old files (information), it takes summaries (photos) and organizes them in a directed acyclic graph structure.

• This method ensures all discussions are preserved indefinitely, allowing easy retrieval of past interactions.

Storage Mechanism

• Information is stored in an SQLite database on the user's computer, which requires no complex setup. Users can modify storage locations through settings.

New Features with Lossless Claw

• Introduces three new functionalities:

• LCM Search: Enables quick searches within conversations similar to using Ctrl + F.

• LCM Describe: Provides summaries of past moments in conversations for quick reference.

• LCM Expand: Allows users to retrieve full details from summarized content.

Installation Requirements

• Requires OpenClaw installation, Node.js version 22 or higher, and an LM provider setup. Installation can be done via command line or GitHub URL.

Configuration Settings

• Recommended environmental variables include:

• LCM Fresh Tail Count: Keeps the last 32 messages detailed for immediate context access.

• LCM Incremental Max Depth: Organizes summaries deeply without limits.

• LCM Context Threshold: Initiates compacting at 75% capacity to maintain response space.

Session Management Tips

• Users can prevent session resets due to inactivity by applying specific configurations; this maintains continuity even with prolonged breaks in interaction.

Additional Tools Available

• A terminal UI (TUI) feature allows visual browsing of the memory database, enabling users to manage summaries effectively—repairing corrupted entries or transferring data between conversations.

Will Summarization Slow Down My AI Agent?

Performance and Efficiency of AI Agents

• The summarization process occurs in the background after each interaction, ensuring that the AI agent operates at full speed without delays.

• Lossless Claw addresses the issue of AI agents forgetting information when their context window is filled by saving every message to a database and summarizing older content into a smart Directed Acyclic Graph (DAG).

• Users can install Lossless Claw with a single command and adjust environmental variables to enhance session management, such as increasing session idle time.

Tools and Features of Lossless Claw

• An optional Text User Interface (TUI) is available for browsing and managing the memory database, which is open-source under the MIT license.

• The AR Profit Boardroom provides comprehensive resources including step-by-step frameworks, use case examples, and 100 prompts for effective implementation of Lossless Claw.

What Are the Community Benefits?

Community Engagement and Support

• The community boasts over 137 pages of testimonials showcasing user success stories within the AI automation space.

• With a membership base of 2,600 individuals, there are ample opportunities for real-time learning and support through live video coaching calls.

Learning Resources Available

• Members have access to extensive training materials designed to take users from beginner to expert in building AI agents quickly.

• Regular updates on new features like browser integration are provided alongside detailed guides on using OpenClaw effectively.

How Can I Stay Updated with New Developments?

Continuous Learning Opportunities

• Daily updates include new courses on various topics such as client acquisition strategies using AI SEO techniques.

• Members can learn about recent advancements in models like Kim K 2.5 Nvidia through practical demonstrations during live sessions.

Exploring New Models

• Discussion around alternative models includes GPT 5.4 and Hunter Alpha, which offers an extensive token context window for free usage via Open Router.

What’s New with OpenClaw?

Latest Updates on Model Integration

• A guide has been created detailing how to set up browser automation with OpenClaw following its latest update featuring Kim K 2.5 Nvidia enhancements.

Practical Applications

• The Gimme K2.5 model allows web search capabilities integrated seamlessly into OpenClaw without requiring API keys or subscriptions.

Using Kim K 2.5 and GLM 5 with OAMA

Overview of New Features

• The latest update allows users to run Kim K 2.5 and GLM 5 models directly with OAMA, utilizing Nvidia's new hardware for enhanced performance.

• The cloud update significantly boosts model performance, providing faster output and responses while integrating free web search capabilities.

Introduction to OAMA

• OAMA enables the running of local and cloud models for free, although users must adhere to token limits.

• Users can easily install Kim K 2.5 through a simple command in their terminal.

Running Commands in Terminal

• A quick command allows users to run Kim K 2.5 cloud directly in the terminal, demonstrating its speed with basic queries.

• Users can also launch Open Claw using a specific command that integrates the new Nvidia model.

Utilizing Clawed Code

• Clawed code can be used alongside Kim K 2.5 cloud for free, showcasing its powerful AI capabilities.

• Interaction with clawed code is notably faster than previous versions, enhancing user experience.

Setting Up GLM 5 Cloud

• To use GLM 5 cloud, users need to execute a command in the terminal while ensuring OAMA is active.

• Installation may take longer due to the size of GLM 5 but offers access to advanced features from Gillette.

Onboarding Process

• Users can onboard through their terminal by executing an installation command for Open Claw.

• This process involves setting base URLs and selecting desired models (cloud or local).

Final Setup Steps

• After resetting configurations in Open Claw, users can interact with the AI using commands like "wake up my friend."

• Successful setup allows interaction with Kim K 2.5 cloud via Open Claw, including web search functionality.

Usage Limits and Considerations

• There are usage limits on cloud tokens that reset frequently; however, regular usage remains feasible within these constraints.

• Users are encouraged to utilize sub-agent functionalities within Open Claw for efficient operations without exceeding limits.

AI Automation and the Latest Developments in AI Models

Overview of Olama's Cloud AI Service

• A company called Olama has updated its free cloud AI service to utilize Nvidia's latest powerful computer chips, specifically the B300.

• The platform allows users to run various AI models on their computers via the cloud, functioning like a free app store for AI capabilities.

Key Features of Kim K2.5 Model

• The Kim K2.5 model, developed by Moonshot AI, is trained on approximately 15 trillion tokens, enabling it to process images and text effectively.

• It features a 256,000 token context window, allowing it to remember extensive conversations akin to reading a small book.

Importance of NVIDIA B300 Chips

• The NVIDIA B300 is designed specifically for running large language models more efficiently and quickly than previous generations.

• This advancement enhances the responsiveness of AI agents, making them faster in providing answers.

Agent Swarm Technology

• Kim K2.5 employs "agent swarm" technology that divides complex tasks into smaller segments handled by multiple mini agents working simultaneously.

• This parallel processing approach improves both speed and accuracy in task completion.

Resources for Learning and Community Support

• A comprehensive guide on using OpenClaw with OAMA is available within the AR Profit community, including video tutorials and setup guides.

• Members can access a supportive community of 2,600 individuals who provide real-time assistance through weekly coaching calls and local meetups.

Educational Opportunities within the Community

• The program offers structured learning paths from beginner to expert levels in AI automation over five weeks.

• Participants receive playbooks detailing effective strategies for automating various business processes such as social media content creation and SEO optimization.