ANOLISA: An Agentic OS Layer That Strips Tool Responses Before They Reach the Model
ANOLISA (Agentic Nexus Operating Layer & Interface System Architecture) | Agentic OS with runtime, security, observability, and Tokenless response compression for lower token usage and cost.
At a glance
- What is it?
- ANOLISA bundles a shell copilot, a response compressor, a memory store and an eBPF tracer into one Apache-2.0 install. The compression path is the part with a concrete mechanism and a number attached; the rest of the stack is broader and less specified in the README.
- Who is it for?
- ANOLISA brings together tools for agent execution, observation, memory and tool-output optimization. Evaluate the individual component and its supported platform before adopting the stack.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly Rust, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- alibaba/anolisa666 ★ · +62
| Repository | 2026-09-11 | 2026-11-15 |
|---|---|---|
| alibaba/anolisa | 604 | 666 |
62 Stars in the last 65 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
Where ANOLISA fits in an agent workload
Tool results and schemas can consume context, but the proportion varies by task. ANOLISA groups components for agent entry, integration and system software. Tokenless targets tool output and schemas, while other components address observation, memory, workspace snapshots and security. Measure your own workload before assuming that output compression will materially reduce cost.
How Token-less compresses a tool response
Tokenless documents tool-output compression, schema simplification and command rewriting. The exact transformations depend on its configuration and version. Compression can remove or alter information an agent needs, so compare representative outputs and task results before rolling it out. The project provides a measurement guide at https://agentic-os.sh/docs/user-guide/tokenless/measuring-savings/; its measurements should not be treated as guaranteed savings for another workload.
Installation and adapter integration
Start with https://agentic-os.sh/docs/user-guide/getting-started/installation/ and the documentation for the components you need. ANOLISA CLI manages component installation, while AW connects agent frameworks and system software through adapters. Requirements and available integration paths change by component and version. Review installation scripts and platform prerequisites before running them, then test your specific framework integration.
AgentSight traces from the kernel with eBPF
AgentSight observes agent trajectories, model calls and token usage. Its Linux implementation uses eBPF for observation, but deployment still depends on the host kernel, permissions and supported runtime. Container boundaries and managed environments may affect visibility. Consult the current AgentSight documentation and validate the signals available in your own environment.
The rest of the stack: checkpoints, sandbox, memory, skills
The current component list includes Agent Memory for persistent memory, SkillFS for skill discovery, workspace checkpoints for snapshots and rollback, and Agent Security Core. AW connects agent frameworks to system software; cosh and OS Skills provide entry-point tools. These descriptions do not establish the security or recovery properties of a particular deployment. Review the documentation and test the components and versions you intend to use.
When to evaluate alternatives
Tokenless may add little value if tool output and schemas represent a small share of your workload. If an agent relies on information that compression removes, its behaviour can change; test representative outputs and retain the original where your workflow requires it. Platform and permission requirements also vary among components, so check them individually before deployment.
How this differs from SDK-based agent platforms
ANOLISA integrates system-level components with existing agent frameworks through adapters and providers. SDK-based tracing and context management can instead be integrated in application code. Either approach can involve setup or infrastructure constraints: compare supported frameworks, operating systems, visibility and maintenance requirements for your deployment. Kernel tracing is one possible way to observe an agent without changing its code, not the only one.
Licence, versions and what maintenance looks like
The repository is Apache-2.0 licensed; review the actual licence and the terms of the individual components and dependencies used in your deployment. Components can have independent releases, so check their current compatibility and platform requirements before upgrading. Read the installer and deployment documentation before running remote installation scripts, and record the versions you test.
Editorial conclusion
ANOLISA brings together tools for agent execution, observation, memory and tool-output optimization. Evaluate the individual component and its supported platform before adopting the stack. For Tokenless, measure the effect on your own workload and verify that compression preserves the information your agent needs; follow the current installation and integration guides for your framework.
Frequently asked questions
What is ANOLISA?
A server-side operating layer for AI agent workloads from Alibaba, written in Rust and Apache-2.0 licensed. It names three practical constraints of agent execution, terminal entry, token cost and execution environments, and the stated aim is to keep the shell, agent framework and sandbox you already use while adding a single CLI entry point whose capabilities enable independently.
How do I install ANOLISA?
Start with the current installation guide at https://agentic-os.sh/docs/user-guide/getting-started/installation/. Supported platforms and installation steps vary by component and integration; check the guide before installing or enabling an adapter.
How does Token-less reduce token usage?
Tokenless reduces token use through techniques including tool-output compression, schema simplification and command rewriting. Its effect and the information retained depend on the configuration and workload; evaluate the results against representative tasks.
What does AgentSight need to run?
AgentSight observes agent trajectories, model calls and token use. Its Linux tracing uses eBPF; check the component's current documentation for kernel, privilege and platform requirements before deployment.
Do all ANOLISA components share one version?
Components are maintained and released separately. Check the versions and compatibility requirements documented for the particular components you plan to combine.
Official sources
Was this useful?