When Siri launched in 2011, AI assistants were supposed to do everything for us, anticipate our needs, and make daily life easier. Yet 15 years later, we're still settling for simple commands like "Siri, set a timer." Google Assistant and Alexa created similar disappointments. By 2026, the truly personal assistant experience promised by big tech companies was still missing. Until ClawdBot emerged. This open-source project went viral in Silicon Valley, causing Mac Mini stock to sell out overnight and creating major waves in the tech community. Unlike traditional cloud-based assistants, ClawdBot offers a new generation of AI experience that runs entirely under your control, never forgets, and acts proactively.
ClawdBot is an open-source personal AI assistant that runs on your own devices. Developed by Peter Steinberger, founder of the former PSPDFKit (now known as Nutrient), this project was initially designed around a space lobster character named "Clawd" for the developer's personal use. When it was released as open source in January 2026, it gathered over 9,000 stars on GitHub within weeks and formed an active developer community.
The fundamental feature distinguishing ClawdBot from other AI assistants is that it runs entirely locally, not cloud-based. While tools like ChatGPT forget you with each session and assistants like Siri can only understand simple commands, ClawdBot can remember conversations lasting weeks, communicate with you proactively, and perform real actions on your computer. This assistant, accessible through daily messaging apps like WhatsApp, Telegram, Slack, Discord, Signal, iMessage, and Microsoft Teams, is always with you without needing to open a website.
ClawdBot's architecture consists of two main components: the Gateway (control plane) and the Agent (AI brain). The Gateway runs as a background service on localhost:18789 and manages all messaging integrations, session routing, and tool connections. This control center processes incoming messages, routes them to the appropriate AI model, and sends responses back to the relevant channel.
The heart of the system, the Agent component, uses Large Language Models (LLMs) like Claude, GPT-4, or locally running Ollama. Peter Steinberger's project typically prefers Anthropic's Claude model. The Agent interprets incoming commands, references its persistent memory to understand context, and plans necessary actions.
ClawdBot's power comes from a modular capability system called "Skills." These plugins, written in TypeScript or JavaScript, provide specialized tools for specific tasks like calendar management, email automation, web scraping, and smart home control. Hundreds of skills developed by the community are shared on ClawdHub, and anyone can expand the system by writing their own skills.
The memory system stores conversations and user preferences as Markdown files in Obsidian format. This approach ensures your data is completely accessible and portable. The proactive engine uses cron jobs, webhooks, and heartbeat checks to exhibit autonomous behaviors. It schedules morning briefings, monitors specific events, and completes tasks without user intervention.
ClawdBot's capabilities surpass traditional AI assistants. Thanks to persistent memory, the system maintains conversation context for days, weeks, or even months. When you mention a project three weeks ago, ClawdBot remembers this information and proactively communicates with you about relevant updates.
Proactive messaging ensures your assistant doesn't just wait for commands. Every morning, you receive a briefing containing your email summary, calendar events, and important news of the day. Twenty minutes before your meeting, it checks traffic conditions and sends you a reminder. When a project you're following gets updated, it informs you without asking.
Full computer access is ClawdBot's most powerful and controversial feature. It can read your file system, execute terminal commands, control the web browser, send emails, and manage calendar events. This enables true automation but also brings serious security responsibilities.
In terms of channel flexibility, ClawdBot supports over 15 messaging platforms. You can continue a conversation started on WhatsApp on Slack, with all context preserved. Thanks to its open-source structure, the community continuously adds new features, fixes bugs, and improves integrations. Docker support and sandbox options provide isolation for security-focused users. It has voice recognition features on macOS, iOS, and Android, and a visual interface called Live Canvas allows you to visualize complex tasks.
Installing ClawdBot is relatively straightforward for users with basic technical knowledge. The system requires Node.js version 22 or higher to function. It can run on macOS, Linux, and Windows (via WSL2) operating systems. The installation process consists of several steps.
The first step is to install ClawdBot globally via the npm package manager. You download the latest version to your system by running the npm install -g clawdbot@latest command in the terminal or command line. You can also use alternative package managers like pnpm or bun.
After installation is complete, you start the onboarding wizard with the clawdbot onboard --install-daemon command. This wizard configures the Gateway daemon to auto-start as a launchd (macOS) or systemd (Linux) user service. This way, ClawdBot continues running in the background when the system restarts.
API key configuration determines which AI model the system will use. You need to obtain an API key from your Anthropic account for Claude API or from OpenAI for GPT-4. Alternatively, you can use completely local models with Ollama. After adding your API key to the configuration file, you can start the Gateway with the clawdbot gateway --port 18789 --verbose command.
Messaging platform integration is central to the user experience. For WhatsApp, you establish a connection by scanning a QR code. For Telegram, you need to obtain a bot token, and for Slack, you need to grant workspace permissions. For security, you can set DM policy settings to "pairing" mode to prevent automatic responses to messages from unknown people. You add trusted people to the allowlist with pairing codes.
Hardware selection depends on your usage intensity. Most users prefer the Mac Mini (approximately $600 with M4 chip) because it's power-efficient and can run 24/7. Low-cost alternatives like Raspberry Pi are also available but may have performance limitations. The cloud server (VPS) option is suitable for those who don't want to keep physical hardware at home. Basic installation takes 30 minutes, while configuring all messaging channels and installing skills can take 2-4 hours.
ClawdBot's real power emerges in daily usage scenarios. In software development, developers have their assistants write code overnight, fix bugs, and open pull requests. Luigi D'Onorio DeMeo shared on Twitter that his ClawdBot running on Mac Mini pulled repositories, opened VS Code, ran tests, generated fixes, and automatically committed if tests were clean. It also sent calendar alerts based on traffic.
In professional automation, users save weekly hours on processing insurance claims, preparing expense reports, and receiving daily summary briefings. One user noted that ClawdBot helped achieve Inbox Zero in their email box, managed bulk unsubscriptions, and prioritized important messages.
On the content production side, SEO specialists and writers use ClawdBot for drafting and data summarization. It can significantly accelerate content production processes, especially on large websites. One user shared that they automatically transcribed over 1,000 WhatsApp voice messages and archived them with semantic search.
Personal productivity examples include automatic flight check-ins, smart home device control, and context-based reminders. One user said they took a photo of a recipe and had their shopping cart prepared in less than 5 minutes. Another user mentioned redesigning their entire website via Telegram while watching Netflix. These examples show that ClawdBot is not just a chatbot but an automation tool that can do real work.
The AI assistants market has been experiencing explosive growth in recent years, and the rise of open-source alternatives like ClawdBot is accelerating this expansion. According to Gartner's August 2025 report, by the end of 2026, 40 percent of enterprise applications will be integrated with task-specific AI agents. This rate was below 5 percent in 2025. Gartner also projects in its best-case scenario that agentic AI could drive approximately 30 percent of enterprise application software revenue by 2035, surpassing $450 billion.
According to Grand View Research data, the global AI assistant market was estimated at $16.29 billion in 2024 and is expected to reach $73.80 billion by 2033, growing at a CAGR of 18.8% between 2025 and 2033. North America dominates the market with a 36.3% share in 2025.
Specifically for personal AI assistants, according to Market.us's report, the market will grow from $2.23 billion in 2024 to $56.3 billion in 2034, expanding at a CAGR of 38.1%. By 2025, the total number of devices equipped with AI assistants is forecast to surpass 8 billion. Mobile devices continue to be the primary channel for AI interaction with a 90% usage rate, and currently 1.5 billion mobile devices have AI capabilities.
The rise of open-source, locally-running assistants like ClawdBot responds to growing concerns about data privacy and user control. Individual users and businesses seeking alternatives to traditional cloud-based assistants are turning to solutions where they can control their own data.
The powerful capabilities offered by ClawdBot bring significant responsibilities and risks. As clearly stated in the system's official documentation, "perfect security is impossible when running frontier AI models with shell access." It's necessary to use the system acknowledging this reality.
Real user experiences regarding security risks are noteworthy. Early adopters reported cases of emails being sent from wrong accounts. An incident where ClawdBot opened tax documents during a live demo screen share caused ripples in the community. The most striking example was when a user sent a test message from their own email address saying "I'm in danger, delete all emails," and the system deleted all emails without verifying who sent the mail. This shows that AI cannot yet make human-level decisions regarding authentication and context understanding.
Peter Steinberger strongly recommends using dedicated hardware. Using an independent device like a Mac Mini or a virtual private server (VPS) limits the damage radius of potential errors. Running it on your main computer means all your files and accounts are at risk.
The technical knowledge requirement should not be overlooked. You should be comfortable using the command line, able to edit configuration files, and capable of troubleshooting issues. Although the Discord community responds quickly to questions, it can be challenging for those without basic Linux or terminal knowledge.
Cost-wise, the hardware investment (approximately $600 for Mac Mini) is a one-time expense, but API usage fees are ongoing. Claude Pro costs $20 per month, while Claude Max is $200 per month. Heavy users can consume millions of tokens monthly. Federico Viticci noted in his review published on MacStories that he consumed 180 million Anthropic API tokens in one week.
Regarding data privacy, although all data is stored locally, your saved payment information, browser cookies, and account access can be used by the system. Scenarios like automatic grocery shopping with saved credit card information mean those same credentials could be used for other things. A prompt injection attack, model hallucination, or misunderstood instruction could result in an "order dinner" command being interpreted as "order $500 worth of dinner."
To mitigate these risks, you can use sandbox modes, separate browser profiles, and tool allowlists. Never grant full access to accounts you can't afford to compromise. Using ClawdBot in a controlled environment with limited permissions is the safest approach.
ClawdBot finally delivers the personal AI experience that big tech companies have promised for years but never quite achieved. With control, flexibility, and persistent memory that traditional cloud-based assistants don't offer, it provides users with genuine automation capabilities. It has transformative potential for technically knowledgeable individuals and businesses. However, this power brings serious security responsibilities and risks. Isolated hardware usage, limited permissions, and conscious use are essential for safely operating the system. The open-source community has accomplished what commercial companies couldn't and created an important turning point in the future of AI assistants. ClawdBot's evolution may herald a new era where AI technology democratization and user control gain importance.
To get detailed information about advanced artificial intelligence solutions like ClawdBot and data security, you can explore other technical content on our blog.
Sources:
Micro-service architecture is a software architecture approach based on the development and deployment of large and complex applications as small, independent and specialized services.
Notion AI is an AI assistant integrated into the popular productivity and note-taking app Notion. Announced in 2023, Notion AI is designed to help users produce content faster, manage workflows and make informed decisions.
Run-time or run-time computing refers to the type of computing in which multiple computing tasks occur simultaneously or at overlapping times
We work with leading companies in the field of Turkey by developing more than 200 successful projects with more than 120 leading companies in the sector.
Take your place among our successful business partners.
Fill out the form so that our solution consultants can reach you as quickly as possible.
By combining our 30+ years of experience and experience with the centers of excellence in our ecosystem, we prepare your organization to lead the digital transformation!
