Simpleyyt/ai-manus

What it solves

AI Manus is a general-purpose AI agent system designed to execute complex tasks by interacting with a real-world environment. It solves the problem of LLMs being limited to text generation by providing a secure, isolated sandbox where the agent can run code, browse the web, and manage files.

How it works

The system uses a "Plan-and-execute" flow. When a user provides a task, a PlanAct Agent creates a structured plan and executes it step-by-step. The agent interacts with a Docker-based Ubuntu sandbox containing a headless Chrome browser and API services for shell and file access.

To allow users to monitor the agent, the system converts VNC streams from the sandbox's browser into WebSockets, enabling real-time viewing and manual takeover of the agent's actions.

Who it’s for

This project is for developers and users who want to deploy their own autonomous AI agent capable of performing actual operations like web research, coding, and file manipulation in a controlled environment.

Highlights

• Isolated Sandboxing: Each task is allocated a separate Docker container to ensure security and isolation.
• Native Tool Use: Uses structured output tools for planning and execution rather than fragile prompt-based JSON protocols.
• Integrated Toolset: Includes built-in support for Terminal, Browser, File, Web Search, and external MCP (Model Context Protocol) tools.
• Real-time Monitoring: Provides a WebUI with the ability to view the agent's browser and interrupt or stop the agent's progress.
• Session Management: Uses MongoDB and Redis to manage session history and support background tasks.