agentic-os-org/ANOLISA

What it solves

ANOLISA provides an operating system layer designed to manage AI agent workloads. It addresses three primary challenges: the need for better terminal-based agent interaction, the high cost of tokens caused by verbose tool outputs, and the lack of robust execution environments for agents.

How it works

ANOLISA acts as a middleware layer that sits between your existing agent frameworks and the underlying system. It provides a CLI to enable specific capabilities, such as:

• Terminal Integration: cosh-ng allows agents to operate directly within a standard shell, blending natural language commands with traditional terminal behavior.
• Token Efficiency: The Token-less component compresses tool schemas and responses before they reach the LLM, significantly reducing token consumption without requiring changes to the agent's code.
• Runtime Security & Observability: It uses tools like eBPF for agent tracing, provides sandboxing for risky operations, and includes checkpointing to allow for workspace rollbacks.

Who it’s for

Developers and engineers who are building or deploying AI agents and need to optimize costs, improve security, or integrate agentic workflows directly into their terminal-based development environments.

Highlights

• Non-intrusive: Works with your existing agent frameworks and sandboxes without requiring code rewrites.
• Token Savings: Demonstrates significant reductions in token usage (e.g., ~60% in tool-heavy pipelines) by stripping redundant data from model context.
• Security Focused: Aligned with AARM standards and maps to the OWASP Agentic Top 10 security risks.
• Observability: Provides deep visibility into agent behavior and token consumption from the kernel level up.