HugAgentOS vs operational-ontology
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 — Minimal TypeScript reference implementation of the Operational Ontology pattern behind Palantir Foundry: shared objects and links, action-gated writes, business rules, audit and write-back.
Same idea at very different weights: hugagentos ships domain ontology as a control plane for agents inside a full enterprise AgentOS, operational-ontology isolates the pattern in a small reference you can read in an afternoon.
| HugAgentOS | operational-ontology | |
|---|---|---|
| Stars | 1.1k | 167 |
| Forks | 66 | 20 |
| Language | Python | TypeScript |
| License | MIT | MIT |
| Last activity | today | today |
| Topics | agents, knowledge-graphs, orchestration | knowledge-graphs, data |
| Curated connections | 10 | 3 |
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.
operational-ontology — the curator's take
Read this before you build or buy an 'ontology' for agents. It draws the line most projects blur: a semantic layer lets an agent read the business, an operational ontology lets it run it, because every write goes through named actions that enforce rules, audit refused attempts and write back to the system of record. The runnable demo (two legacy order systems merged, a shipped order that refuses cancellation) makes the pattern concrete in code you can fork. It says itself it is a learning resource, not a framework: no package, no scaling story, one maintainer. Use it to design the action layer your agents need; do not deploy it.