HugAgentOS vs utopia
ZJU's self-hosted enterprise AgentOS on AgentScope 2.0: domain ontology as a control plane for agents, plus RAG, sub-agents, MCP, skills, sandbox, memory and approval-gated self-evolution. — versus — DeepLethe's open 'enterprise world model': one Rust binary plus Postgres running a bitemporal knowledge graph with ontology packs, cited hybrid search, an agent harness and MCP. Air-gap ready.
Both make an ontology-governed knowledge graph the enterprise substrate for agents; utopia is the bitemporal world model with cited retrieval, HugAgentOS the whole AgentOS wrapped around one.
| HugAgentOS | utopia | |
|---|---|---|
| Stars | 1.1k | 8.0k |
| Forks | 63 | 1.0k |
| Language | Python | Rust |
| License | NOASSERTION | Apache-2.0 |
| Last activity | yesterday | 2 days ago |
| Topics | agents, knowledge-graphs, orchestration | knowledge-graphs, rag |
| Curated connections | 9 | 6 |
HugAgentOS — the curator's take
Pick HugAgentOS when the pitch 'ontology as a machine-executable control plane' matters to you - governed concepts, relations, rules and action contracts feed the skill, memory and orchestration engines one shared business vocabulary, and self-evolution (memory, skills, orchestration) only lands after you approve it. One-command install with SQLite, Docker Compose for Postgres/Redis and a real sandbox. Caveats: the repo is a generated mirror of an upstream (`src/**` PRs not accepted), the licence is Apache-2.0 *plus* supplementary terms, and the Community Edition has no self-registration. If you don't care about ontology governance, maxkb or claraverse are more mature self-hosted workspaces; if you only want the ontology layer, context-ontology-accelerator or semantica are lighter.
utopia — the curator's take
For a company that wants a governed knowledge base its agents can trust, on its own hardware: ingest PDFs/Office/Markdown and sync sources (GitHub, Jira, Notion, S3), extraction follows an ontology you pick at creation (schema.org, W3C Org, PROV-O, FOAF, IOF ship in the binary), facts are bitemporal so you can ask what was known as of a date, and a built-in agent walks the graph with the same read-only tools exposed over MCP. Tantivy full-text is embedded, vectors go in pgvector, the job queue is a table — nothing else to run. NOT a memory layer for an agent app (it is a product, not a library) and not for a single person's notes: the ontology and audit machinery only pay off when several people and agents share the knowledge.