A powerful, extensible framework for building AI agents with comprehensive tool support, multi-agent orchestration, and advanced web automation capabilities.

ManusUse empowers developers to create sophisticated AI agents that can:

• Execute code in secure Docker sandboxes
• Automate web browsing and data extraction
• Analyze data and generate visualizations
• Coordinate multiple specialized agents for complex tasks
• Integrate with various LLM providers seamlessly

Built on Strands SDK and integrated with browser-use, ManusUse provides a production-ready foundation for AI agent development.

• ManusAgent: General-purpose agent for file operations and code execution
• BrowserUseAgent: Advanced web automation using natural language
• DataAnalysisAgent: Specialized for data processing and visualization
• VulnerabilityIntelligenceAgent: Deep CVE analysis with multi-source intelligence (va_agent.py)
• VulnerabilityDiscoveryAgent: Automated vulnerability discovery and submission (vd_agent.py)
• MCPAgent: Model Context Protocol integration for extended capabilities
• WorkflowAgent: Complex task orchestration with dependency management

• File operations (read, write, edit, delete)
• Code execution in Docker sandboxes
• Web search and content retrieval
• Browser automation (click, type, extract)
• Data visualization (charts, plots, reports)
• Security analysis tools (CVE checking, threat intelligence)

• Automatic task decomposition and routing
• Parallel execution of independent tasks
• Intelligent agent selection based on task requirements
• Result aggregation and error handling

• Support for OpenAI, Anthropic, AWS Bedrock, Ollama
• Model Context Protocol (MCP) compatibility
• Extensible tool system
• Configuration-driven architecture

# Basic installation
pip install manus-use

# With browser automation support
pip install manus-use[browser]
playwright install chromium

# Full installation with all features
pip install manus-use[browser,search,visualization]

from manus_use import ManusAgent

# Create an agent
agent = ManusAgent(model="gpt-4o")

# Execute a task
response = agent("Create a Python script that fetches weather data and saves it to a CSV file")
print(response)

from manus_use.agents import BrowserUseAgent

# Create a browser agent
browser_agent = BrowserUseAgent()

# Automate web tasks
result = browser_agent("Go to GitHub and find the top 5 trending Python repositories today")

from manus_use.multi_agents import WorkflowAgent

# Create a workflow agent
workflow = WorkflowAgent()

# Execute a complex task with multiple agents
result = workflow.handle_request("""
    1. Search the web for recent AI research papers
    2. Analyze the trends and create visualizations
    3. Generate a comprehensive report with insights
""")

Create a config/config.toml file:

[llm]
provider = "openai"  # or "anthropic", "bedrock", "ollama"
model = "gpt-4o"
api_key = "your-api-key"  # or use environment variable

[sandbox]
enabled = true
docker_image = "python:3.12-slim"

[tools]
enabled = ["file_ops", "code_execute", "web_search", "browser"]

See config.example.toml for all available options.

ManusUse provides multiple CLI interfaces:

# Basic CLI
manus-use "Create a factorial function in Python"

# Enhanced CLI with rich UI
manus-use-enhanced

# Automatic multi-agent mode for complex tasks
manus-use "Research quantum computing applications and create a presentation"

# Clone the repository
git clone https://github.com/manus-use/manus-use.git
cd manus-use

# Install in development mode
pip install -e ".[dev,browser,search,visualization]"

# Run tests
hatch run test

# Check code quality
hatch run lint
hatch run format

# Run all tests
hatch run test

# Run with coverage
hatch run test-cov

# Run specific test
hatch run pytest tests/test_agents.py -v

Explore the examples/ directory for:

• Basic agent usage
• Browser automation
• Multi-agent workflows
• Data analysis

• Quick Start Guide
• Browser Agent Setup
• API Reference

ManusUse includes sophisticated vulnerability intelligence capabilities through specialized agents:

A comprehensive CVE analysis tool that performs deep vulnerability research:

# Analyze a specific CVE
python va_agent.py CVE-2025-6554

# The agent will automatically:
# - Fetch data from NVD and GitHub advisories
# - Check CISA KEV and AlienVault OTX
# - Search for public exploits and PoCs
# - Perform deep PoC analysis and validation
# - Generate a comprehensive Lark document report

Features:

• Multi-source intelligence gathering (NVD, CISA KEV, OTX, GitHub)
• Automatic PoC discovery and validation
• Deep code analysis of exploits
• Threat intelligence correlation
• Automated report generation

An automated vulnerability discovery and submission system:

# Run automated vulnerability discovery
python vd_agent.py

# The agent will:
# - Calculate time slices for the current two weeks
# - Discover CVEs with high EPSS scores
# - Submit vulnerabilities in batches
# - Provide detailed submission summaries

Features:

• Automated time-based CVE discovery
• EPSS score filtering for high-impact vulnerabilities
• Concurrent processing for efficiency
• Batch submission capabilities
• Multi-agent orchestration

# Using the Vulnerability Analysis Agent programmatically
from va_agent import VulnerabilityIntelligenceAgent

agent = VulnerabilityIntelligenceAgent(model_name="us.anthropic.claude-sonnet-4-20250514-v1:0")
result = agent.handle_request("Analyze CVE-2025-6554 and create a comprehensive report")

# Using the Vulnerability Discovery Agent
from vd_agent import VulnerabilityDiscoveryAgent

discovery = VulnerabilityDiscoveryAgent(model_name="us.anthropic.claude-sonnet-4-20250514-v1:0")
result = discovery.handle_request("Execute vulnerability discovery workflow")

Important: These tools are designed for defensive security purposes only and should be used for legitimate security research, vulnerability management, and defense.

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

This project is licensed under the MIT License - see LICENSE for details.

• Built with Strands SDK - A powerful Python SDK for building AI agents
• Browser automation powered by browser-use - Framework for AI-driven web automation
• Inspired by Anthropic's computer use demonstrations

• 📖 Documentation
• 🐛 Issue Tracker
• 💬 Discussions