kernell-os-sdk 1.0.0

Kernell OS Open-Source Framework for Autonomous PC Agents with shared memory, M2M commerce ($KERN), cluster orchestration, and self-evolution. Beats Claude Managed Agents.

Kernell OS SDK

Kernell OS SDK is a production-grade, security-first framework for building autonomous AI agents that can interact with the machine, participate in M2M (Machine-to-Machine) commerce, and operate securely within isolated environments.

Features

• 🛡️ Secure By Default: Hardened sandbox execution, cryptographic identities bound to hardware UDID, and strict permission boundaries (anti-RCE).
• 💸 Built-in M2M Commerce: Native integration with the Kernell Agent Protocol (KAP) and $KERN token for task payments and escrow.
• 🧠 Cortex Shared Memory: Offloads context to Redis for massive token savings using episodic memory streams.
• 📈 Command Center: A beautiful, real-time local dashboard to monitor metrics, toggle permissions, and manage API keys safely.
• 🔋 Production Resilience: Includes Token Budgeting, Circuit Breakers, SLO Monitoring, and Tool Output Persistence out of the box.

Installation

# Core SDK
pip install kernell-os-sdk

# With Dashboard / GUI support
pip install kernell-os-sdk[gui]

# With Redis memory support
pip install kernell-os-sdk[memory]

# Everything
pip install kernell-os-sdk[all]

Quick Start

1. Create a Secure Agent

from kernell_os_sdk import Agent, AgentPermissions

# Create an agent with specific permissions
agent = Agent(
    name="Data Analyst",
    permissions=AgentPermissions(
        network_access=True,
        file_system_read=True,
        execute_commands=False  # Secure by default
    )
)

# Start the agent daemon
agent.run()

2. Launch the Command Center

Monitor your agent, toggle permissions in real-time, and manage API keys without restarting:

from kernell_os_sdk import CommandCenter

# Attach dashboard to your agent
dashboard = CommandCenter(agent, port=8500)
dashboard.start()

# Dashboard available at http://127.0.0.1:8500

3. Add Custom Skills (Tools)

Easily extend your agent with type-hinted skills. API keys are safely retrieved from the Command Center at runtime.

import httpx

@agent.skill(description="Fetch the weather for a city.")
def get_weather(city: str) -> str:
    # Safely get API key added via the Dashboard
    api_key = dashboard.get_api_key("weather_api")
    if not api_key:
        return "Error: Weather API key not configured."
        
    response = httpx.get(f"https://api.weather.com/v1/{city}?key={api_key}")
    return response.text

Advanced Features

Token Budgeting & Circuit Breakers

Protect your API credits and prevent cascading failures:

# Agent has built-in budget tracking
if agent.budget.can_spend(estimated_tokens=500):
    # Safe to call LLM
    pass
else:
    print(f"Budget exhausted! Throttled: {agent.budget.snapshot().throttle_reason}")

M2M Commerce ($KERN)

Pay other agents or receive payments for completed tasks:

from kernell_os_sdk import Wallet

with Wallet() as wallet:
    balance = wallet.get_balance()
    print(f"Current Balance: {balance} KERN")
    
    # Hold funds in escrow
    escrow_id = wallet.request_payment_escrow(amount=5.0, task_id="analyze_data", payer_id="agent_123")
    
    # Release on success
    wallet.release_escrow(escrow_id)

Security Posture

Kernell OS SDK is designed for zero-trust environments:

• Private Keys: Encrypted at rest using AES-128-CBC and bound to the machine's hardware UDID. Passports cannot be cloned.
• Sandboxing: Containerized execution drops all Linux capabilities, prevents root mounting, and enforces disk quotas.
• Command Execution: No shell=True. Commands are strictly sanitized against a blacklist of destructive operations (rm -rf, mkfs, etc.).
• Audit Logging: All permission changes and API key accesses are logged immutably in the dashboard.

Contributing

We welcome contributions! Please see our Contributing Guide for details on submitting pull requests, writing tests, and clean code standards.

# Run tests
pytest tests/ -v

License

MIT License. See LICENSE for details.