ocbot_agent is the core AI extension component of the ocbot project, running as a Chrome Extension. It provides an interactive interface via the Side Panel and leverages low-level browser APIs to achieve automation.
This project is included as a git submodule in both ocbot. You typically don't need to clone it separately — use git clone --recursive on the parent project instead.
- Multi-Model Support: Built-in support for major LLMs such as OpenAI, Anthropic, Gemini, DeepSeek, Moonshot, and more.
- Native Browser Integration: Deeply integrated via the Side Panel, providing a seamless browsing experience without interruption.
- Autonomous Agent: Utilizes
debuggerandscriptingpermissions to read web content and perform automated actions. - Privacy First: Your data remains in your control. Supports local models via compatible interfaces.
- Framework: WXT (Web Extension Tools)
- UI Library: React + Tailwind CSS
- Icons: Lucide React
- Build Tool: Vite (via WXT)
- Node.js >= 18
- npm or pnpm
npm installnpm run devThis command starts a development server and attempts to automatically open Chrome with the extension loaded.
npm run buildBuild artifacts will be located in the .output/ directory.
ocbot_agent/
├── entrypoints/ # Extension entry points (background, sidepanel, content)
├── lib/ # Core logic library
│ ├── agent/ # AI Agent core logic (act, observe, loop)
│ ├── llm/ # LLM adapters
│ └── channels/ # Communication channels (Telegram, etc.)
├── assets/ # Static assets
└── wxt.config.ts # WXT configuration file
This extension requests the following sensitive permissions to implement Agent functionality:
sidePanel: Provides the sidebar interactive interface.debugger: Used for high-privilege browser automation operations.scripting: Used to inject scripts into pages.tabs&activeTab: Used to retrieve tab information.storage: Used to save configuration and chat history.
Pull Requests and Issues are welcome to improve the ocbot agent.
This project is licensed under the MIT License - see the LICENSE file for details.
