The honest comparison
The Urfael assistant, Hermes Agent and OpenClaw: every win and every gap in one table. Urfael wins where it counts for a machine that lives on your desk and acts for you: the smallest blast radius, a flat bill within your plan’s limits, and the ability to prove what it did. We are smaller and newer, and we say so in the same table.
| Capability | Urfael Assistant | Hermes | OpenClaw |
|---|---|---|---|
| No inbound network port | none on the daemon; optional services loopback-only | varies | gateway/DMs |
| Ships an attack benchmark | yes, npm run security | no | no |
| Flat-rate cost (no per-token) | yes, on your Claude Code plan (its limits and terms apply) | no | no |
| Can prove what it did (ledger + seal) | yes | no | no |
| Live, watchable multi-agent Council | yes (experimental) | see their docs | see their docs |
| Undoable coding turns (checkpoint + rewind) | yes, auto-checkpoint + reversible rewind | none documented | none documented |
| Skill hub supply-chain checks | scanned + previewed + sha-pinned | external standard | third-party registry; see their docs |
| Proactive memory recall (every turn) | retrieves per turn | frozen snapshot | agent must search |
| Generative UI that can't run code | sanitized canvas (validator shipped; live rendering not built yet) | no | renders agent UI |
| Chat-channel breadth | 19 (11 native + 8 webhook) | many | 20+ |
| Battle-tested at scale | small, and we say so | large | very large |
Where it counts for a machine on your desk: blast radius, cost, provability, and not overstating maturity. OpenClaw and Hermes are genuinely excellent at breadth and scale.
Want the caveats, not just the wins? See what's lightly tested