ZJU-REAL/HugAgentOS

HugAgentOS: The Self-Evolving AgentOS for Ontology-Grounded Trustworthy Reasoning

What it solves

HugAgentOS addresses the gap between simple AI chat wrappers and functional enterprise agents. It provides a self-hosted foundation that gives AI models the necessary context, execution tools, and structured governance (via domain ontology) to perform real-world tasks—such as running code, managing files, and executing automated workflows—while ensuring data remains private and actions are trustworthy.

How it works

It operates as an "AgentOS" that uses a domain ontology as a machine-executable control plane to govern reasoning and actions. The system employs a ReAct orchestration loop to coordinate between Model Context Protocol (MCP) servers, custom agent skills, and sandboxed execution environments. It supports a three-layer memory system (relational, vector, and graph) and a private RAG-based knowledge base to maintain long-term context. To ensure reliability, candidate plans are passed through deterministic rule checks and evidence reviews before execution.

Who it’s for

It is designed for individuals and enterprises who need a self-hosted, model-agnostic AI agent workspace that can handle complex tasks with gated execution and strict data privacy.

Highlights

  • Ontology-Grounded Reasoning: Uses domain ontologies to provide a shared business vocabulary for skills, memory, and orchestration.
  • Extensible Tooling: Integrates MCP tools and standardized "Agent Skills" for expanded capabilities.
  • Gated Execution: Implements risk-tiered evidence review and rule checks to prevent unauthorized or incorrect actions.
  • Comprehensive Workspace: Includes a data canvas for structured data editing, a lightweight sandbox for code execution, and support for sub-agents.
  • Self-Evolving: Can distill memory and skills from real work, which are then applied only after human approval.

Related

  • Project
    java-up-up/nexus-agent

    An enterprise-grade AI agent platform that implements a complete closed-loop of document governance, hybrid RAG retrieval, and ReAct-based autonomous agents for production environments.

  • Project
    xpert-ai/xpert

    An open-source Agent platform that uses a hybrid agent-workflow architecture and a unified plugin system to build governed, enterprise-grade agentic applications.

  • Project
    framerslab/agentos

    A TypeScript AI agent framework featuring persistent cognitive memory, runtime tool forging, and personality-driven orchestration across 11 LLM providers.

  • Project
    Q00/ouroboros

    Ouroboros is an open‑source *Agent OS* that adds a specification‑first, replayable workflow to any code‑generating LLM (Claude, Codex, Gemini, Copilot, etc.). A one‑line installer provides the `ooo` CLI; you run `ooo setup` then `ooo interview "…"` to have the tool ask Socratic questions, lock the resulting spec (Seed), generate code, automatically evaluate it, and repeat until the contract is satisfied. All actions are recorded in an immutable ledger, making the process auditable and model‑agnostic. The core kernel lives in this repository, while a terminal UI (`ourocode`) and domain‑specific plugins live in companion repos.

  • Project
    PhyAgentOS/PhyAgentOS-core

    An agent framework for embodied AI that uses a decoupled execution gateway and evidence-based verification to ensure robots achieve their goals and recursively improve from experience.