Clawdbot is making waves as a new open-source personal AI assistant that you can run on your own hardware. This innovative tool connects large language models from companies like Anthropic and OpenAI to various practical applications, including messaging apps, files, and smart home devices. The best part? You maintain control over the orchestration layer, keeping your data safe.
What sets Clawdbot apart isn’t just its ability to chat. It offers a clear architecture for local-first agents and features a unique workflow engine called Lobster. This engine allows users to create deterministic pipelines that turn model calls into reliable processes.
At the heart of Clawdbot is the Gateway process. It provides a WebSocket control plane and a local HTTP interface for a user-friendly control UI and web chat. Messages from popular platforms like WhatsApp, Telegram, and Discord are sent to the Gateway, which determines which agent will respond, what tools can be used, and which model provider to call. Replies are then sent back through the same channel.
The system is built around several key concepts. The Gateway handles routing, model calls, and sessions. Nodes allow Clawdbot to access local resources, such as file systems and browser automation. Channels integrate with various chat systems, and Skills are tools that agents can call, described in a standardized format. This structure enables users to run the Gateway on a low-cost virtual server or a spare computer while using powerful model computing on remote APIs when needed.
Clawdbot also introduces an open skills format with its SKILL.md standard. Skills are defined in Markdown, outlining specific procedures for tasks like deployment. The Gateway reads these definitions, making them available to agents as tools with clear capabilities and safety measures. This means operational runbooks can be transformed from informal documents into machine-executable skills.
Another exciting feature is Lobster, the workflow runtime that powers many of Clawdbot’s advanced automations. Lobster allows for multi-step tool sequences to be executed as single operations, enforcing time limits and sandbox policies. This structured approach ensures that workflows remain predictable and auditable.
Clawdbot’s proactive behavior is another reason it’s gaining attention. It can run scheduled jobs and track state across sessions, offering features like daily briefings and monitors that alert you when specific conditions are met. All of this functionality runs securely on your machine or server, while model calls are still made to external providers.
Setting up Clawdbot is straightforward. A simple one-line installer fetches the necessary components, or users can install it via npm or clone the TypeScript repository. After onboarding, users can connect their preferred channels, select a model provider, and enable skills. They can also create their own SKILL.md files and build Lobster workflows.
The excitement around Clawdbot is palpable, with users sharing their experiences on social media. From deploying websites to managing local models, Clawdbot is proving to be a versatile tool for personal AI tasks. As more people discover its capabilities, it’s clear that Clawdbot is at the forefront of the next wave of AI assistants.