A powerful Multi-Agent AI Framework for intelligent automation and orchestration
Moonlight transforms how you build AI applications by providing a comprehensive framework that orchestrates multiple AI agents, automates complex workflows, and integrates with thousands of tools seamlessly.
🏗️ Multiple Execution Modes
🔌 Massive Tool Ecosystem
⚡ Production Ready
# Install
uv pip install moonlight-ai
# Install Playwright (NEEDED)
playwright install
Note: Install NPX and UVX on your machine to get the most out of MCP Compatibility.
import os
from moonlight_ai import Agent, Hive, MoonlightProvider
from dotenv import load_dotenv
load_dotenv()
# Create a content creation provider
provider = MoonlightProvider(
provider_name="deepseek",
api_key=os.getenv("DEEPSEEK_API_KEY"),
)
# Create a specialized content agent
content_agent = Agent(
name="content_creator",
instruction="""You are an expert content creator specializing in technical blog posts.
Create engaging, well-structured content with clear explanations and practical examples.""",
provider=provider,
model_name="deepseek-chat"
)
# Generate a technical blog post
response = Hive(content_agent).run("""
Write a 500-word blog post about the benefits of microservices architecture,
including real-world examples and potential challenges.
""")
print(response)
# Schema for financial analysis
financial_schema = {
"type": "object",
"properties": {
"risk_level": {
"type": "string",
"enum": ["low", "medium", "high"],
"description": "Investment risk assessment"
},
"recommendation": {
"type": "string",
"enum": ["buy", "hold", "sell"],
"description": "Trading recommendation"
},
"confidence": {
"type": "number",
"minimum": 0,
"maximum": 1,
"description": "Confidence score (0-1)"
},
"key_factors": {
"type": "array",
"items": {"type": "string"},
"description": "Key factors influencing the decision"
}
},
"required": ["risk_level", "recommendation", "confidence", "key_factors"]
}
# Create financial analysis agent
analyst_agent = Agent(
name="financial_analyst",
instruction="""You are a senior financial analyst. Analyze stocks and provide
structured recommendations based on market data, company fundamentals, and trends.""",
provider=provider,
model_name="deepseek-chat",
json_mode=True,
json_schema=financial_schema
)
# Analyze a stock
analysis = Hive(analyst_agent).run("""
Analyze Tesla (TSLA) stock based on recent performance, EV market trends,
and competition. Consider Q4 2024 earnings and 2025 outlook.
""")
print(analysis)
# Returns structured JSON with risk_level, recommendation, confidence, and key_factors
from moonlight_ai import MCPConfig, MCPRegistry
# Create MCP configurations for time and weather
time_config = MCPConfig(
name="time",
command="uvx",
args=["mcp-server-time", "--local-timezone=Asia/Kolkata"],
description="Provides current time in IST timezone",
tags=["time", "utility", "timezone"]
)
weather_config = MCPConfig(
name="weather",
command="uvx",
args=["mcp-server-weather", "--api-key=${WEATHER_API_KEY}"],
description="Real-time weather data and forecasts",
tags=["weather", "forecast", "api"]
)
# Create a personal assistant with multiple MCP servers
assistant_agent = Agent(
name="personal_assistant",
instruction="""You are a helpful personal assistant with access to time and weather data.
Provide contextual responses that combine multiple data sources for better insights.""",
provider=provider,
model_name="deepseek-chat",
mcp_servers=[time_config, weather_config]
)
# Get contextual information
response = Hive(assistant_agent).run("""
What's the current time in Mumbai? Also check the weather and let me know
if it's a good time for an outdoor meeting this afternoon.
""")
print(response.assistant_message)
registry = MCPRegistry()
# Register a server
reistry.register(time_config)
# load a config
registry.get_by_name("time") # or search by tags based on the sdk
from moonlight_ai import Orchestra, Agent
# Create specialized research agents
market_researcher_agent = Agent(
name="market_researcher",
instruction="""You are a market research specialist. Focus on industry trends,
market size, growth rates, and competitive landscape analysis.""",
provider=provider,
model_name="deepseek-chat",
)
competitor_analyst_agent = Agent(
name="competitor_analyst",
instruction="""You are a competitive intelligence expert. Analyze competitor
strategies, pricing, product features, and market positioning.""",
provider=provider,
model_name="deepseek-chat",
)
financial_analyst_agent = Agent(
name="financial_analyst",
instruction="""You are a financial analyst specializing in startup and tech company
valuations, funding rounds, and financial performance metrics.""",
provider=provider,
model_name="deepseek-chat",
)
# Set up the research orchestra
research_team = {
"market_researcher_agent": market_researcher_agent,
"competitor_analyst_agent": competitor_analyst_agent,
"financial_analyst_agent": financial_analyst_agent
}
orchestrator = Orchestra(
orchestra_model="deepseek-chat",
provider=provider
)
orchestrator.add_agents(research_team)
orchestrator.add_task("""
Conduct a comprehensive analysis of the AI code assistant market in 2025.
I need insights on:
1. Market size and growth projections
2. Key competitors and their positioning
3. Recent funding rounds and valuations
4. Emerging trends and opportunities
Provide a structured report with actionable insights for a startup entering this space.
""")
# Execute the research
results = orchestrator.activate()
print(results)
from moonlight_ai import DeepResearcher
# Create a deep researcher for emerging technologies
tech_researcher = DeepResearcher(
model="deepseek-chat",
provider=provider,
research_task="""
Analyze the current state and future prospects of quantum computing in 2025.
Focus on: commercial applications, major players, recent breakthroughs,
investment trends, and timeline for mainstream adoption.
""",
max_depth=3, # Deep recursive research
breadth=4, # Multiple angles per level
links_per_query=8, # Comprehensive source coverage
max_content_length=2000 # Detailed content analysis
)
# Execute comprehensive research
print("🔍 Starting deep research on quantum computing...")
results = tech_researcher.run()
# Get the detailed research report
print("📄 Research Report:")
print("=" * 80)
print(results)
from moonlight_ai import WorkflowGenerator, Workflow, Agent
# Create specialized trading agents
market_scanner = Agent(
name="market_scanner",
instruction="""You are a market scanning specialist. Monitor stock prices,
volume, and technical indicators. Identify potential trading opportunities.""",
provider=provider,
model_name="deepseek-chat"
)
risk_manager = Agent(
name="risk_manager",
instruction="""You are a risk management expert. Evaluate trading opportunities
for risk/reward ratios, position sizing, and portfolio impact.""",
provider=provider,
model_name="deepseek-chat"
)
trade_executor = Agent(
name="trade_executor",
instruction="""You are a trade execution specialist. Execute trades based on
approved signals while managing order timing and execution quality.""",
provider=provider,
model_name="deepseek-chat"
)
available_agents = {
"market_scanner": market_scanner,
"risk_manager": risk_manager,
"trade_executor": trade_executor
}
# Generate automated trading workflow
wfg = WorkflowGenerator(
available_agents=available_agents,
model_name="deepseek-chat",
provider=provider,
)
# Create trading workflow
trading_prompt = """
Create a trading workflow that:
1. Scans for stocks breaking above 20-day moving average with high volume
2. Analyzes risk/reward ratio and position sizing
3. Executes trades only if risk/reward > 2:1
4. Sets stop-loss at 2% and take-profit at 4%
5. Runs every 15 minutes during market hours
"""
print('🤖 Generating automated trading workflow...')
# Generate and execute the workflow
# Step 1: Generate first level workflow
print('\n1. Generating first level workflow...')
first_level_result = wfg.generate_first_level(trading_prompt)
print(f'First level result: {first_level_result}')
# Step 2: Expand first workflow (simulate critique and requirements)
# You should check the first level, then answer the requirements and then expand the workflow based on the critique.
print('\n2. Expanding first workflow...')
# Simulating a critique for the first level workflow
critique = "The workflow needs to be more specific about time checking and should include proper error handling."
# Assuming we have some answered requirements based on the critique
answered_requirements = [
"Q. xxx; A. Check current minute value",
"Q. xxx; A. Determine if minute is even or odd",
"Q. xxx; A. Print appropriate message"
]
previous_level_details = first_level_result
expanded_workflow = wfg.expand_first_workflow(
critique=critique,
answered_requirements=answered_requirements,
previous_level_details=previous_level_details
)
print(f'Expanded workflow: {expanded_workflow}')
# Step 3: Generate final workflow
print('\n3. Generating final workflow...')
final_workflow = wfg.generate_final_workflow(expanded_workflow)
# Final Workflow can be a json or generated.
print(f'Final generated workflow: {final_workflow}')
print('\n' + '=' * 60)
# Create and run the trading workflow
trading_workflow = Workflow(
workflow=final_workflow,
available_agents=available_agents,
provider=provider,
model_name="deepseek-chat"
)
print('📈 Starting automated trading workflow...')
trading_workflow.run()
Perfect for focused tasks requiring deep expertise:
Best For: Content creation, data analysis, specialized consulting
Ideal for complex tasks requiring diverse expertise:
Best For: Research projects, business analysis, multi-step processes
Specialized for comprehensive research workflows:
Best For: Market research, academic investigation, competitive analysis
Advanced automation for complex business processes:
Best For: Business process automation, monitoring systems, trading bots
Moonlight includes validated configurations for thousands of MCP servers across categories:
from moonlight_ai import MCPRegistry, MCPConfig
# Create and register configurations
registry = MCPRegistry()
# Database configuration
db_config = MCPConfig(
name="postgres",
command="uvx",
args=["mcp-server-postgres", "--connection-string=${DB_URL}"],
description="PostgreSQL database integration",
tags=["database", "sql", "postgres"]
)
registry.register(db_config)
# Retrieve and use saved configurations
saved_config = registry.get_by_name("postgres")
database_agent = Agent(
name="db_analyst",
instruction="You are a database analyst with PostgreSQL access.",
provider=provider,
model_name="deepseek-chat",
mcp_servers=[saved_config]
)
Moonlight supports all major AI providers with unified interfaces:
# Switch between providers seamlessly
openai_provider = MoonlightProvider(
provider_name="openai",
api_key=openai_key
)
deepseek_provider = MoonlightProvider(
"deepseek",
api_key=deepseek_key
)
groq_provider = MoonlightProvider(
"groq",
api_key=groq_key
)
custom_provider = MooonlightProvider(
provider_url="https://my_custom_server/v1",
api_key=custom_key
)
# Same agent, different backends
agent_gpt4 = Agent(
"analyst",
instruction="...",
provider=openai_provider,
model_name="gpt-4"
)
agent_deepseek = Agent(
"analyst",
instruction="...",
provider=deepseek_provider,
model_name="deepseek-chat"
)
Made with 🌙 by the Moonlight Team
FAQs
🌙 Advanced Multi-Agent AI Orchestration Framework
We found that moonlight-ai demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Following last week’s supply chain attack, Nx published findings on the GitHub Actions exploit and moved npm publishing to Trusted Publishers.
Security News
AGENTS.md is a fast-growing open format giving AI coding agents a shared, predictable way to understand project setup, style, and workflows.
Security News
/Research
Malicious npm package impersonates Nodemailer and drains wallets by hijacking crypto transactions across multiple blockchains.