Agent OS
The Linux Kernel for AI Agents
Current AI frameworks let LLMs decide everything—including whether to follow safety rules. Agent OS inverts this: the kernel decides, the LLM just computes. Zero policy violations. Guaranteed.
$ pip install agent-os-kernel
Why Agent OS?
The difference between hoping your agent behaves and guaranteeing it.
The Problem: Hope-Based Safety
system_prompt = """
You are a helpful assistant.
Please don't delete important files.
Be careful with sensitive data.
Always ask before taking action.
"""
# 🤞 Hope the LLM complies...The Solution: Kernel Enforcement
from agent_os import KernelSpace
kernel = KernelSpace(policy="strict")
@kernel.register
async def my_agent(task):
# Kernel intercepts every action
# Blocked actions never execute
return await process(task)The OS Analogy
Agent OS applies proven operating system principles to AI governance.
Core Modules
Each module solves a specific problem and can be used independently.
CMVK
Cross-Model Verification Kernel. Detect hallucinations by comparing outputs across multiple LLMs.
Learn more →EMK
Episodic Memory Kernel. Immutable agent memory with semantic search and time-travel debugging.
Learn more →IATP
Inter-Agent Trust Protocol. Cryptographic identity and message signing for multi-agent systems.
Learn more →AMB
Agent Message Bus. Decoupled communication with Redis, Kafka, NATS, and SQS backends.
Learn more →Policy Engine
Define what agents can and cannot do. Policies enforced at kernel level—before actions execute.
Learn more →POSIX Signals
Control execution with SIGKILL (terminate), SIGSTOP (pause), and SIGCONT (resume).
Learn more →Framework Integrations
Wrap your existing agents with one line. No rewrites required.
🦜 LangChain
from agent_os.integrations import langchain_kernel
governed = langchain_kernel.wrap(agent)👥 CrewAI
from agent_os.integrations import crewai_kernel
governed = crewai_kernel.wrap(crew)🤖 AutoGen
from agent_os.integrations import autogen_kernel
governed = autogen_kernel.wrap(assistant)🔮 OpenAI Assistants
from agent_os.integrations import openai_kernel
governed = openai_kernel.wrap(assistant)Production Use Cases
Agent OS in action across regulated industries.
Healthcare
HIPAA-compliant AI agents with guaranteed PHI protection.
0 PHI incidentsFinance
Risk-aware agents with position limits and compliance guardrails.
$2.3M risk preventedEducation
Safe tutoring bots with age-appropriate content and parent visibility.
COPPA compliantEnterprise
Multi-agent coordination with cryptographic trust.
SOC 2 readyGet Started in 2 Minutes
Install
pip install agent-os-kernelCreate a Governed Agent
from agent_os import KernelSpace
kernel = KernelSpace(policy="strict")
@kernel.register
async def my_agent(task: str):
# Your agent logic here
# Kernel enforces policies automatically
return f"Completed: {task}"Execute Safely
import asyncio
result = asyncio.run(kernel.execute(my_agent, "analyze data"))
print(result) # "Completed: analyze data"📚 Tutorials
Step-by-step guides to master Agent OS, from basics to production deployment.
Your First Governed Agent
Learn the fundamentals of Agent OS by building a simple governed agent from scratch.
Detecting Hallucinations with CMVK
Use multi-model consensus to verify AI outputs and catch hallucinations before they cause problems.
Building Multi-Agent Systems
Coordinate multiple agents with cryptographic trust using IATP and the Agent Message Bus.
📖 Documentation
Everything you need to build governed AI agents.