Estimated tool-schema tokens reduced by 47.3% in the reference test.
Tokenless · one Schema fixture · estimated tokens. See the measurement
System Software Stack for AI Agents.
Context. Skills. Recovery. Security. Observability.
The users of operating systems have changed. ANOLISA makes Agents first-class citizens.
Install the ANOLISA CLI
curl -fsSL https://get.agentic-os.sh | bash01 / CAPABILITY DEMOS
Interactive examples with sample data. No commands are executed.
COSH-NG · TERMINAL COLLABORATION
Keep your terminal. Keep your session.
Let Agents work in your familiar Shell. Take control at any time.
Natural-language tasks, failure insights, and login health checks.Open the guide ↗
Terminal sessioncosh-ng
InteractionSHELL + AGENT
❯Create a Vite project; I will answer interactive prompts
npm create vite@latest admin-demo?Install with npm and start now?Yes / No
The command is waiting for your input
02 / WHY ANOLISA
Better systems. Better Agents.
Enhance the Agent applications, frameworks, and harnesses you already use.
Keep context useful.
Compress tool output and reuse memory across sessions.
Find the right Skills.
Discover OS instructions and Skills when a task needs them.
Keep control. Roll back changes.
Constrain actions with policies. Inspect and restore workspace snapshots.
See what happened.
Inspect model and tool calls, trajectories, and Token usage.
Build it into your ecosystem.
OS vendors & distribution communities
Integrate ANOLISA into an existing distribution, or include it in your Agentic OS design from day one. Adopt and help build the stack.
Agent governance platforms & sandbox providers
Preinstall ANOLISA in your runtime environments to give Agent workloads system support.
03 / THE SYSTEM SOFTWARE STACK
Three layers. One stack.
From user interaction to unified integration and Agent Semantic Layer capabilities.
3.1 / User Entrypoint Layer
3.2 / AW Unified Stack Layer
3.3 / Agent Semantic Layer
AgentSight
Observe Agent execution through trajectories, LLM calls, and Token usage.
Agent Security Core
OS-level security kernel for AI Agents.
Tokenless
Reduce AI Agent Token usage through tool output compression, tool schema simplification, command rewriting, and other techniques.
Agent Memory
Give Agents searchable, persistent memory to retain and reuse information across sessions.
SkillFS
Organize Skills into stable filesystem views for Agents to discover and read on demand.
Workspace Checkpoint
Create Agent workspace snapshots, compare file differences, and roll back changes when needed.
Installation and maintenance: ANOLISA CLI manages component installation and lifecycle. ktuner: Inspect Linux kernel parameters, get tuning recommendations, and apply or roll back changes.
04 / OPEN ARCHITECTURE
Extend the stack through AW.
Connect your Agent, or bring your own system software optimized for Agent workloads.
Agent integration
Connect Agents and frameworks to AW.
Provider development
Bring your system software into ANOLISA.
SDK
Call component capabilities from application code.
05 / COMMUNITY
Build ANOLISA with us.
Contribute
Improve the code, refine the docs, or share an image or cloud service entry.
Stay connected
Share what you build and help shape what comes next.
Agent Security Core references: AARM Aligned · OWASP Agentic Top 10 control mapping