skyrelis
Advanced tools
AI Agent Security Library - Enterprise-grade security for AI agents, starting with comprehensive observability across multiple frameworks
Enterprise-grade security for AI agents, starting with comprehensive observability.
As AI agents become more powerful and autonomous, they present new security challenges:
Skyrelis provides the security foundation your AI agents need.
🔍 Complete Observability - Full visibility into agent execution and decision-making
🎯 System Prompt Security - Monitor and protect agent instructions and behaviors
📊 Real-time Monitoring - Instant alerts for suspicious agent activities
🏷️ Agent Registry - Centralized inventory and security posture management
🔗 Zero-Config Integration - Add security with just a decorator
⚡ Production Ready - Built for enterprise scale and reliability
🌐 Standards Compliant - OpenTelemetry, audit logging, and compliance ready
🚀 Multi-Framework Support - LangChain (0.1-2.0), CrewAI (0.70+), and extensible architecture
✅ Modern LangChain Compatible - Full support for LangChain 1.0+ with inheritance-based monitoring
🛡️ Prompt Injection Detection - AI-powered input validation and threat detection
🏗️ Agent Sandboxing - Isolated execution environments with controlled permissions
👥 Access Control & RBAC - Role-based permissions for agent operations
🧠 Behavioral Analysis - ML-based anomaly detection for agent activities
📋 Compliance Frameworks - SOC2, GDPR, HIPAA compliance tools
🔐 Secret Management - Secure handling of API keys and sensitive data
# Basic installation
pip install skyrelis
# With CrewAI support
pip install skyrelis[crewai]
# With all features
pip install skyrelis[all]
Skyrelis supports multiple AI agent frameworks with unified security monitoring:
from skyrelis import observe_langchain_agent
from langchain_core.runnables import Runnable
from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage, SystemMessage
# Modern LangChain agent with Skyrelis monitoring
@observe_langchain_agent(remote_observer_url="https://your-security-monitor.com")
class ModernSecureAgent(Runnable):
def __init__(self, llm_model="gpt-4o-mini"):
self.llm = ChatOpenAI(model=llm_model)
def invoke(self, input_data, config=None, **kwargs):
messages = [
SystemMessage(content="You are a helpful AI assistant."),
HumanMessage(content=input_data["query"])
]
return self.llm.invoke(messages)
# Use your secure agent
agent = ModernSecureAgent()
result = agent.invoke({"query": "What's the weather like?"})
from skyrelis import observe_langchain_agent
from langchain.agents import AgentExecutor, create_openai_functions_agent
from langchain_openai import ChatOpenAI
from langchain.prompts import ChatPromptTemplate
# Legacy LangChain agent setup
prompt = ChatPromptTemplate.from_messages([
("system", "You are a helpful AI assistant. Use tools when needed."),
("human", "{input}"),
("placeholder", "{agent_scratchpad}")
])
llm = ChatOpenAI(model="gpt-4o-mini")
agent = create_openai_functions_agent(llm, tools, prompt)
# Add enterprise security monitoring with one decorator! 🔒
@observe_langchain_agent(remote_observer_url="https://your-security-monitor.com")
class SecureAgent(AgentExecutor):
pass
# Initialize and use - now with full security monitoring
secure_agent = SecureAgent(agent=agent, tools=tools)
result = secure_agent.invoke({"input": "What's the weather like?"})
from skyrelis import observe_crewai_agent
from crewai import Agent, Task, Crew
@observe_crewai_agent(remote_observer_url="https://your-security-monitor.com")
class SecureCrewAgent(Agent):
pass
# Your CrewAI agent now has complete security monitoring
agent = SecureCrewAgent(
role="Security Analyst",
goal="Analyze security threats",
backstory="Expert in cybersecurity analysis"
)
All supported frameworks automatically get:
When you add the @observe decorator, Skyrelis automatically captures security-relevant data:
@observe(
monitor_url="https://your-security-monitor.com",
agent_name="customer_service_agent",
security_level="production", # "development", "staging", "production"
)
class CustomerServiceAgent(AgentExecutor):
pass
@observe(
monitor_url="https://your-security-monitor.com",
agent_name="financial_advisor_agent",
security_level="production",
enable_audit_logging=True, # Full audit trail
enable_anomaly_detection=True, # Behavioral analysis (coming soon)
enable_input_validation=True, # Prompt injection detection (coming soon)
compliance_mode="SOC2", # Compliance framework (coming soon)
alert_thresholds={ # Security alerting
"unusual_tool_usage": 0.8,
"response_time_anomaly": 2.0,
"error_rate_spike": 0.1
}
)
class FinancialAdvisorAgent(AgentExecutor):
pass
# Security monitoring endpoints
export SKYRELIS_MONITOR_URL="https://your-security-monitor.com"
export SKYRELIS_SECURITY_LEVEL="production"
# Compliance and audit
export SKYRELIS_AUDIT_RETENTION_DAYS="2555" # 7 years for financial compliance
export SKYRELIS_COMPLIANCE_MODE="SOC2"
# Alert destinations
export SKYRELIS_SLACK_WEBHOOK="https://hooks.slack.com/..."
export SKYRELIS_SECURITY_EMAIL="security-team@company.com"
from skyrelis import observe
from langchain.agents import create_openai_functions_agent
from langchain_openai import ChatOpenAI
from langchain.tools import StructuredTool
def get_account_balance(account_id: str) -> str:
# This tool access is now fully monitored and audited
return f"Account {account_id}: $10,000"
@observe(
monitor_url="https://security.bank.com/monitor",
security_level="production",
compliance_mode="SOX",
enable_audit_logging=True
)
class BankingAgent(AgentExecutor):
pass
# Every interaction is now compliance-ready and security-monitored
@observe(
monitor_url="https://security.company.com/monitor",
enable_anomaly_detection=True, # Detect unusual customer behavior
enable_input_validation=True, # Block prompt injection attempts
alert_on_threats=True # Real-time security alerts
)
class CustomerServiceAgent(AgentExecutor):
pass
# Agent automatically detects and blocks security threats
@observe(
monitor_url="https://security.research.com/monitor",
data_classification="confidential",
enable_data_loss_prevention=True, # Prevent sensitive data exposure
audit_data_access=True # Log all data access events
)
class ResearchAgent(AgentExecutor):
pass
# Complete data protection and access monitoring
The Skyrelis Security Monitor provides:
Skyrelis Security Architecture:
All security monitoring happens transparently - your agent code remains unchanged while gaining enterprise-grade security!
✅ RESOLVED: 'method' object attribute '__init__' is read-only Error
Previous versions of Skyrelis had compatibility issues with LangChain 1.0+ due to class protection mechanisms. This is now fixed!
__init__ method assignment failed in modern LangChainObservedAgent classes# OLD APPROACH (Failed in LangChain 1.0+)
cls.__init__ = new_init # ❌ Read-only error
# NEW APPROACH (Works with all LangChain versions)
class ObservedAgent(cls): # ✅ Inheritance-based
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
# Add Skyrelis monitoring
No code changes needed! Your existing Skyrelis decorators work with both:
# Basic security monitoring (supports all LangChain versions)
pip install skyrelis
# With CrewAI support
pip install skyrelis[crewai]
# With OpenTelemetry integration
pip install skyrelis[opentelemetry]
# With advanced security features (coming soon)
pip install skyrelis[security]
# With compliance reporting
pip install skyrelis[compliance]
# With threat detection (coming soon)
pip install skyrelis[threat-detection]
# Everything
pip install skyrelis[all]
'method' object attribute '__init__' is read-only error in LangChain 1.0+We welcome contributions to make AI agents more secure! Please see our Contributing Guide for details.
Skyrelis is proprietary software - see the LICENSE file for details.
📧 Licensing Inquiries: security@skyrelis.com
As an AI agent security platform, Skyrelis requires:
Made with 🔒 by the Skyrelis Security Team
Skyrelis: Securing AI agents for the enterprise.
FAQs
AI Agent Security Library - Enterprise-grade security for AI agents, starting with comprehensive observability across multiple frameworks
We found that skyrelis 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
OpenAI rotated macOS signing certificates after a malicious Axios package reached its CI pipeline in a broader software supply chain attack.
Security News
Open source is under attack because of how much value it creates. It has been the foundation of every major software innovation for the last three decades. This is not the time to walk away from it.
Security News
Socket CEO Feross Aboukhadijeh breaks down how North Korea hijacked Axios and what it means for the future of software supply chain security.