SEC EDGAR agentkit for smolagents

A lightweight integration of SEC EDGAR data access tools for Hugging Face's smolagents framework.

Installation

pip install sec-edgar-agentkit-smolagents
# or
pip install -r requirements.txt

Quick Start

from sec_edgar_smolagents import create_sec_edgar_agent

# Create an agent with all SEC EDGAR tools
agent = create_sec_edgar_agent("gpt-4")

# Ask questions about companies and filings
result = agent.run("What was Apple's revenue last year?")
print(result)

Available Tools

• CIKLookupTool - Look up company CIK by name or ticker
• CompanyInfoTool - Get detailed company information
• CompanyFactsTool - Retrieve XBRL company facts
• FilingSearchTool - Search for SEC filings
• FilingContentTool - Extract filing content
• Analyze8KTool - Analyze 8-K material events
• FinancialStatementsTool - Extract financial statements
• XBRLParseTool - Parse XBRL data
• InsiderTradingTool - Analyze Forms 3/4/5

Usage Examples

Basic Company Research

from sec_edgar_smolagents import create_sec_edgar_agent

agent = create_sec_edgar_agent("gpt-4")

# Simple queries
agent.run("Find Microsoft's latest 10-K filing")
agent.run("What were Tesla's material events in the last quarter?")
agent.run("Show me insider trading activity for Apple")

Using Specific Tools

from sec_edgar_smolagents import CIKLookupTool, FilingSearchTool

# Create agent with only specific tools
agent = create_sec_edgar_agent(
    "gpt-4",
    tools=[CIKLookupTool(), FilingSearchTool()]
)

result = agent.run("Find Amazon's CIK and recent filings")

Different Model Providers

# OpenAI
agent = create_sec_edgar_agent("gpt-4")

# Anthropic
agent = create_sec_edgar_agent("claude-3-opus-20240229")

# Hugging Face API
from smolagents import HfApiModel
model = HfApiModel("meta-llama/Llama-3-8b-instruct")
agent = create_sec_edgar_agent(model)

# Local model
from smolagents import TransformersModel
model = TransformersModel("microsoft/Phi-3-mini-4k-instruct")
agent = create_sec_edgar_agent(model)

Financial Analysis

agent = create_sec_edgar_agent("gpt-4")

# Complex multi-step analysis
analysis = agent.run("""
    1. Find Apple's last 3 years of 10-K filings
    2. Extract revenue and net income from each
    3. Calculate year-over-year growth rates
    4. Compare with Microsoft's performance
""")

print(analysis)

MCP Server Configuration

The tools connect to the sec-edgar-mcp server. By default, it looks for the server at the sec-edgar-mcp command. You can customize this:

from sec_edgar_smolagents import SECEdgarToolkit
from sec_edgar_smolagents.mcp_client import MCPClient

# Custom MCP server command
client = MCPClient(server_command="/path/to/sec-edgar-mcp")
toolkit = SECEdgarToolkit(mcp_client=client)

agent = create_sec_edgar_agent("gpt-4", tools=toolkit.get_tools())

Development

Running Tests

pytest integrations/smolagents/__tests__/

Running Examples

cd integrations/smolagents
python examples/basic_usage.py
python examples/financial_analysis.py

Requirements

• Python 3.8+
• smolagents >= 0.1.0
• sec-edgar-mcp server installed (pip install sec-edgar-mcp)

License

AGPL-3.0 - See LICENSE for details.

Author

Stefano Amorelli stefano@amorelli.tech