OpenClaw Overview (pre-installed on Rose)
realvco pre-installs OpenClaw (open-source AI agent software, built on the Claude Agent SDK) in the Rose (OC-1) container.
For deep OpenClaw learning, see the official manual:
- 📘 OpenClaw official documentation: https://docs.openclaw.ai/
- 💻 OpenClaw GitHub: https://github.com/openclaw/openclaw
This page only briefly introduces OpenClaw’s role on the realvco platform + our specific configuration. All general OpenClaw knowledge (commands, APIs, advanced usage) defers to the official documentation.
OpenClaw in one sentence
OpenClaw = open-source AI agent software built on Claude Agent SDK. It enables AI to do more than chat:
- Multi-turn reasoning (remembers prior conversation)
- Tool calling (Web search, run commands, write files)
- Cross-session memory
- Custom persona (.md config files)
For detailed concepts, see OpenClaw official documentation.
OpenClaw’s role on the realvco platform
Every realvco mVPS pre-installs OpenClaw in the Rose (OC-1) container:
| Container | Default role | Default permission | Feature |
|---|---|---|---|
| OC-1 (Rose) | All-purpose manager | Host Root (72hr) | Cross-container coordination, file ops, deep reasoning |
realvco-specific configurations
The following are added by realvco on top of OpenClaw (not in upstream):
1. Pre-installed + auto-launched
OpenClaw open-source version requires you to git clone + docker-compose up + manage dependencies. On realvco it’s already running — just log into admin-panel.
2. Dashboard (admin-panel) UI
You don’t touch SSH / Linux commands. Via Dashboard you can:
- View AI live status
- Pause / resume AI
- Edit persona / switch model
- Track token usage
- Upgrade software version
3. ROOT access duration
OpenClaw upstream has no “duration” concept. realvco adds ROOT Access Control — 6hr / 24hr / 72hr / 30d, auto-revokes on expiry.
4. Upgrade risk control (realvco shields you from upstream churn)
OpenClaw upstream development is very active — a new release roughly every week or two. New versions occasionally regress (API changes, edge-case bugs). realvco’s admin-panel mechanisms let you control the upgrade pace and recover fast if a version misbehaves, so you don’t have to ride upstream daily:
- You decide when to upgrade — upgrades are triggered manually in admin-panel, never auto-pushed
- Daily 05 auto backup of conversations & settings (7-day rolling); plus a “Backup Now” button anytime
- Local Backup Versions — one-click rollback within 5 days if an upgrade fails (5-10 min to revert)
- Create Restore Point — one-click whole-host snapshot before big changes; worst-case time machine back to before
Recommended: “Backup Now” + “Create Restore Point” before upgrading, watch for 5 days after, then relax or roll back. Operational detail is on the linked admin-panel pages.
5. Platform connection integration
OpenClaw upstream supports webhooks but you’d handle SSL certs, URL routing, token auth yourself. realvco bundles these into Platform Connection — paste token and go.
6. Usage billing integration
realvco provides Usage sub-tab + Token & cost concepts, translating OpenClaw API calls into USD costs you can read.
7. Four pre-installed models + quick aliases
Once you’ve bought pre-installed API credit, realvco installs 4 models in Rose’s OpenClaw container ahead of time. Type /model <alias> mid-chat to switch models instantly:
| Alias | Model | Role | Best for |
|---|---|---|---|
gm | OpenAI GPT-5.4 Mini | Primary (default) | Daily steady workhorse / settings & adjustments / config & env checks / general OpenClaw / Hermes-Agent operations |
ds | DeepSeek V4 Pro | Fallback 1 | Heavy tasks & long context / complex logs / multi-step debugging |
hk | Anthropic Claude Haiku 4.5 | Fallback 2 | High-stakes gatekeeping / Claude-style steady judgment |
gf | Google Gemini 3.5 Flash | Fallback 3 | Multi-modal & high-tier backup / more expensive / final option |
Example: while chatting with Rose, type /model ds and send — the next reply uses DeepSeek V4 Pro. Type /model gm to switch back to GPT-5.4 Mini.
Which one to switch to? Pick by task type — see Model escalation rule.
This is OpenClaw’s upstream model alias mechanism. When you later swap the pre-installed key for your own OpenRouter key (or eventually move to OpenAI / Anthropic direct), all 4 aliases carry over — aliases are not bound to a specific key. Complete command syntax, adding your own aliases, and swapping models beyond the 4 defaults — see docs.openclaw.ai Persona / Models section.
Customers on the older default model set: early installs shipped with a different 4-model set (
km/mm/sn/op, now-retired models). The current default isgm/ds/hk/gf— to move to the new set, re-pull the default in Version Upgrade; otherwise you can keep the old set running.
OpenClaw use cases (conceptual)
OpenClaw is suited for:
- Complex reasoning: customer questions need multi-turn thinking
- Cross-tool tasks: AI calling multiple tools to complete
- Long conversations: customer service, deep consultation
- Custom persona: AI talks with specific style
Specific case studies and examples: see OpenClaw official documentation Examples / Use Cases section.
vs Hermes-Agent
See What Rose, Ada and Vi Are Each Good At.
In a line:
- OpenClaw (Rose) = deep reasoning, cross-tool chaining, brand voice
- Hermes-Agent (Ada / Vi) = fast, low cost, platform breadth, self-maintaining skill library
Rose comes with OpenClaw pre-installed, Ada / Vi with Hermes-Agent, all managed uniformly by realvco.
What to learn deeper
realvco operation layer (in this docs site):
OpenClaw general knowledge (go to upstream docs):
- Tool syntax → docs.openclaw.ai
- Agent SDK API → GitHub
- Advanced persona design (SOUL / IDENTITY details) → docs.openclaw.ai
- Skills / MCP integration → docs.openclaw.ai
Related
- Hermes-Agent Overview (pre-installed on Ada / Vi) — what Ada / Vi run
- What Rose, Ada and Vi Are Each Good At
- Persona setup — How to configure on realvco
- OpenClaw release notes — Updates per version on realvco