{"slug": "anolisa-agentic-os-with-runtime-security-observability-and-token-compression", "title": "Anolisa – Agentic OS with runtime, security, observability and token compression", "summary": "Alibaba released ANOLISA, an open-source agentic operating system layer for AI Agent workloads, featuring a terminal interface (cosh-ng), token compression (Token-less), observability (AgentSight), and runtime security. In one observed coding task, Token-less saved 317K tokens (40.5%) based on AgentSight measurements, with tool responses reduced by 65.8% and tool schemas by 47.3%.", "body_md": "**A** gentic **N** exus **O** perating **L** ayer & **I** nterface **S** ystem **A** rchitecture\n\n**The operating system layer for Agent workloads.**\n\nLet Agents drive the system straight from your terminal, and strip the tool responses that reach the model before they cost you — while keeping the Shell, Agent framework, and sandbox you already run.\n\n[中文版](/alibaba/anolisa/blob/main/README_zh.md) · [Website](https://agentic-os.sh/) ·\n[Quick Start](https://agentic-os.sh/docs/quickstart/) ·\n[User Guide](https://agentic-os.sh/docs/user-guide/) ·\n[Contributing](https://github.com/alibaba/anolisa/blob/main/CONTRIBUTING.md)\n\nANOLISA is a server-side operating layer for AI Agent workloads. It addresses three practical constraints of Agent execution: terminal entry, Token cost, and execution environments. Keep the Shell, Agent framework, and sandbox you already use. ANOLISA CLI provides a single installation entry point, while each capability can be enabled independently.\n\n**New to ANOLISA?**\n[Choose your first outcome in the Quick Start →](https://agentic-os.sh/docs/quickstart/)\n\n| Agent entry | Context efficiency | Runtime & security |\n|---|---|---|\nShell copilot |\nTool-output compression |\nCheckpoint and rollback |\nSystem and DevOps expertise |\nTrace and Token visibility |\nFocused Skill views |\nKernel tuning |\nCross-session memory |\nSandbox and verification |\nSandbox lifecycle |\n\n**01 · AGENT INTERFACE**\n\ncosh-ng is an AI-native Linux terminal: it keeps familiar Bash/Zsh behavior, then adds an Agent that can understand intent, use tools and Skills, and ask for approval before risky work. Shell commands and natural language share one terminal instead of forcing users into a separate chat application.\n\n**02 · CONTEXT EFFICIENCY**\n\n[Token-less](https://agentic-os.sh/docs/user-guide/token-saving/tokenless/quickstart/)\nremoves redundancy from tool schemas and responses before they reach the model.\n[Agent Memory](https://agentic-os.sh/docs/user-guide/token-saving/agent-memory/)\nreuses useful context across sessions.\n[SkillFS](https://agentic-os.sh/docs/user-guide/runtime/skillfs/) keeps the\ncurrent Skill view focused and makes other Skills discoverable when needed.\n[AgentSight](https://agentic-os.sh/docs/user-guide/agent-observability/agentsight/)\nshows where Tokens are spent.\n\nOn Linux, AgentSight uses eBPF to observe an Agent without changing its code. Follow user input through model and tool calls, with Token use and sub-agent branches in the same view.\n\n## agentsight-demo-en-ebpf-cover-1080p.mp4 |\n\nInstall Token-less and connect it to Claude Code:\n\n```\ncurl -fsSL https://get.agentic-os.sh | bash\nexport PATH=\"$HOME/.local/bin:$PATH\"\nanolisa install tokenless\nanolisa adapter enable tokenless claude-code\n```\n\nRestart Claude Code, run one tool-heavy task, then inspect the result:\n\n```\ntokenless stats summary\ntokenless stats list --limit 5\n```\n\n[Open the full Token-less Quick Start →](https://agentic-os.sh/docs/user-guide/token-saving/tokenless/quickstart/)\n· [Read the user manual](https://agentic-os.sh/docs/user-guide/token-saving/tokenless/user-manual/)\n\n## tokenless-context-efficiency-hd.mp4 |\n\nIn one observed coding task, Token-less saved 317K Tokens (40.5%), based\non AgentSight measurements.\nResults vary by workload.\n\n`debug`\n\nand `trace`\n\nare dropped by the field blacklist, `metadata`\n\nas null, and\n`tags`\n\n/ `extra`\n\nas empty values. Compression runs between the Agent and the\nmodel, so no Agent framework code changes. Dropped array items stay retrievable\nthrough a `<<tokenless:KEY>>`\n\nmarker, which keeps the compression reversible.\n\n| Tool responses | Tool schemas | Full pipeline |\n|---|---|---|\n65.8% fewer Tokens |\n47.3% fewer Tokens |\n62.9% fewer Tokens |\n| ResponseCompressor · 46.85 µs | SchemaCompressor · 11.44 µs | 198.91 µs |\n\nSavings apply to the tool responses entering the context, not to the whole\nsession bill. The [Token-less user manual](https://agentic-os.sh/docs/user-guide/token-saving/tokenless/user-manual/)\nexplains how to estimate the effect for a given workload.\n\n**03 · EXECUTION RUNTIME**\n\nANOLISA is building out the Agent execution environment:\n[Agent Sec Core](https://agentic-os.sh/docs/user-guide/agent-security/agent-sec-core/quickstart/)\nisolates risky operations, and\n[ws-ckpt](https://agentic-os.sh/docs/user-guide/runtime/ws-ckpt/) keeps recovery\npoints for workspace changes.\n\nWhen a signed Skill changes, the Agent reports `drifted`\n\nbefore using it again.\nA rescan records blocking findings as `deny`\n\n.\n\n[Try the Agent demo →](https://agentic-os.sh/docs/user-guide/agent-security/agent-sec-core/qoder-skill-ledger-demo/)\n· [Skill Ledger guide](https://agentic-os.sh/docs/user-guide/agent-security/agent-sec-core/skill-ledger/)\n\n## skill-ledger-demo-en-with-cover.mp4 |\n\n[Choose a runtime or security starting point →](https://agentic-os.sh/docs/quickstart/)\n· [Start with ANOLISA CLI](https://agentic-os.sh/docs/user-guide/user-entrypoint/anolisa-cli/)\n\nANOLISA CLI is the common installation entry point. cosh-ng is installed in system mode; Token-less and other capabilities can be added independently.\n\n```\ncurl -fsSL https://get.agentic-os.sh | bash\n\nsudo anolisa --install-mode system install cosh-ng\nanolisa install tokenless\n```\n\nRun `cosh`\n\nto enter the AI-native terminal. Token-less can also optimize tool\ncalls from an existing Agent without changing its framework.\n\n[Quick Start](https://agentic-os.sh/docs/quickstart/) ·\n[Installation](https://agentic-os.sh/docs/user-guide/installation/) ·\n[User Guide](https://agentic-os.sh/docs/user-guide/) ·\n[Troubleshooting](https://agentic-os.sh/docs/user-guide/troubleshooting/) ·\n[Build from Source](https://agentic-os.sh/docs/building/) ·\n[Changelog](https://agentic-os.sh/changelog/)\n\n[Open an issue](https://github.com/alibaba/anolisa/issues)for bugs and feature requests.- Read\n[CONTRIBUTING.md](https://github.com/alibaba/anolisa/blob/main/CONTRIBUTING.md)before submitting a pull request. - Report vulnerabilities through the\n[Security Policy](https://github.com/alibaba/anolisa/blob/main/SECURITY.md).\n\nANOLISA is released under the\n[Apache License 2.0](https://github.com/alibaba/anolisa/blob/main/LICENSE).", "url": "https://wpnews.pro/news/anolisa-agentic-os-with-runtime-security-observability-and-token-compression", "canonical_source": "https://github.com/alibaba/anolisa", "published_at": "2026-08-21 03:36:01+00:00", "updated_at": "2026-08-21 03:43:14.377978+00:00", "lang": "en", "topics": ["ai-infrastructure", "ai-tools", "ai-agents"], "entities": ["Alibaba", "ANOLISA", "cosh-ng", "Token-less", "AgentSight", "Claude Code", "SkillFS"], "alternates": {"html": "https://wpnews.pro/news/anolisa-agentic-os-with-runtime-security-observability-and-token-compression", "markdown": "https://wpnews.pro/news/anolisa-agentic-os-with-runtime-security-observability-and-token-compression.md", "text": "https://wpnews.pro/news/anolisa-agentic-os-with-runtime-security-observability-and-token-compression.txt", "jsonld": "https://wpnews.pro/news/anolisa-agentic-os-with-runtime-security-observability-and-token-compression.jsonld"}}