Overview
What is AI Agents in Container?
A guide to set up a containerized Linux desktop (Kasm) running browser-use and web-ui for AI agents, along with an MCP Puppeteer server for web automation. Aimed at developers who want a reproducible environment for browser-based AI agents.
How to use AI Agents in Container?
Follow the provided Docker commands to launch a Kasm desktop container, then inside it clone and install browser-use and web-ui. For MCP Puppeteer, build the Docker image and paste the given JSON configuration into VS Code Insiders, Claude Desktop, Cursor, or Codium Windsurf.
Key features of AI Agents in Container
- Kasm Ubuntu Noble Desktop with VNC access on port 8080
- browser-use and web-ui installed inside the container
- Ollama integration for local LLM models (e.g., deepseek-r1:8b)
- MCP Puppeteer server configured for multiple IDEs/clients
- Support for Google AI Studio API keys
- Setup instructions for both Linux (Docker) and macOS
Use cases of AI Agents in Container
- Run browser-based AI agents in an isolated, reproducible container environment
- Use MCP Puppeteer to give AI coding assistants web browsing capabilities
- Prototype and test browser-use agents with a clean desktop UI
- Combine local LLMs (via Ollama) with web automation in a single container
FAQ from AI Agents in Container
What browser-based tools are included?
browser-use and web-ui are installed. The Kasm desktop provides a graphical browser environment.
How do I install and run Ollama?
Run the provided shell commands inside the container for Linux, or download the macOS installer. Then pull a model like deepseek-r1:8b.
How do I set up the MCP Puppeteer server?
Build the Docker image from the MCP servers repository, then add the given JSON configuration to your editor’s MCP settings (VS Code Insiders, Claude Desktop, Cursor, or Codium Windsurf).
What dependencies are required?
Docker (for container), Python 3, UV package manager, Playwright, and optionally Ollama. The README provides full installation steps.
How do I access the web-ui after setup?
Open http://localhost:7788 in a browser (or the appropriate host address). For the container, forward ports 6901 (VNC) and 7788 (web-ui).
