Meet the 3 AI agents
This page covers the role positioning + pre-installed software differences of the 3 agents on the realvco platform. AI agent advanced usage defers to the software’s upstream documentation:
- 📘 OpenClaw (Rose) → docs.openclaw.ai
- 📘 Hermes-Agent (Ada / Vi) → see linked resources in Hermes-Agent release notes per version
After subscribing realvco, you get 3 AI agents, default roles:
| Agent | Container | Pre-installed software | Default Name | Role |
|---|---|---|---|---|
| 🌹 | OC-1 | OpenClaw | Rose | All-purpose manager, persistent Host Root |
| 🎨 | HM-2 | Hermes-Agent | Ada | High-volume support, creative writing, knowledge Q&A |
| 🔬 | HM-3 | Hermes-Agent | Vi | Platform breadth, English scenarios, self-maintaining skill library |
Why 3?
Unlike ChatGPT / Claude where “one session = one AI”, realvco gives you 3 AI containers by default, because:
- Division of labor: Rose handles deep reasoning and system management, Ada takes high-volume support, Vi covers multi-platform traffic — three independent flows, no interference
- Isolation: One AI breaks (persona corrupted, attacked) → other two still running
- Trial and error: Test new prompts on one companion without affecting the other two’s real work
- Multi-platform: Rose handles Telegram, Ada handles LINE, Vi handles Discord — each manages its own
These are all pre-installed by realvco: the Rose container runs OpenClaw; the Ada and Vi containers run Hermes-Agent; all wrapped in realvco’s custom management interface — nothing for you to install. Each one’s strengths are described below.
You don’t have to use all 3 — using just one is fine; idle ones don’t burn tokens.
Rose (OC-1) — All-purpose manager
| Property | Setting |
|---|---|
| Pre-installed software | OpenClaw (open-source AI agent software, on Claude Agent SDK) |
| Default permission | Host Root (72 hours, ROOT duration adjustable) |
| Default model | claude-sonnet-4-6 / opus-4-7 |
| Best for | System management, cross-container coordination, file ops, complex reasoning |
Rose is the only agent with Host Root enabled by default. This means Rose can touch host files, install software, run system commands — default 72-hour grant, auto-revoked on expiry.
If you don’t need this capability (most customers don’t), you can change it in Settings → ROOT Access.
Ada (HM-2) — High-volume support, creative writing, knowledge Q&A
| Property | Setting |
|---|---|
| Pre-installed software | Hermes-Agent (open-source AI agent software; same as Vi) |
| Default permission | None (container isolated) |
| Default model | See pre-installed model list on the costs page |
| Best for | Customer service, creative writing, knowledge Q&A, English audiences, high-volume daily conversation |
Ada is pre-installed with Hermes-Agent (same as Vi) — fast replies, low token cost, strong English. A natural fit for high-volume customer support and everyday conversation.
When you need deep reasoning, cross-tool chaining, or long conversations that preserve context, hand the task to Rose. Most customers still route daily work to Ada (Rose stays for management tasks and heavy reasoning).
Vi (HM-3) — Platform breadth, English scenarios, self-maintaining skill library
| Property | Setting |
|---|---|
| Pre-installed software | Hermes-Agent (same as Ada) |
| Default permission | None |
| Default model | See pre-installed model list on the costs page |
| Strengths | 22 messaging platforms, self-maintaining skill library, fast replies, cost efficiency |
Vi runs the same Hermes-Agent as Ada — platform breadth is Vi’s signature: the messaging platform list keeps growing (now 22, including Microsoft Teams, Tencent Yuanbao, Google Chat, LINE, SimpleX Chat), and the skill library tidies itself automatically (Curator runs in the background every 7 days).
When to route tasks to Vi:
- Need to reach many messaging platforms (Vi’s platform breadth is the strongest)
- Want a self-maintaining skill library (one less thing to manage by hand)
- High-volume simple customer questions (FAQ-level) — Ada works too; pick Vi when you want to split traffic with Ada
- English-user-dominated scenarios
Which one should I use?
| Your need | Recommended |
|---|---|
| 24/7 customer service bot (Chinese / English) | Ada (HM-2) |
| Need AI to write scripts, run system management | Rose (OC-1) + ROOT |
| Need to reach many messaging platforms | Vi (HM-3) |
| Long conversations / deep reasoning / cross-tool chaining | Rose (OC-1) |
| Deep copywriting / brand-voice polishing | Rose (OC-1) |
| Cross-tool automation (cron + webhook) | Rose (needs ROOT) |
| I just need one | Ada (handles volume; the default pick) |
Can the 3 agents collaborate?
Yes, but requires setup. See Multi-agent collaboration.
By default the 3 agents are independent (don’t know what each other is doing). To enable collaboration:
- In one agent’s persona setup, tell them “you can call Ada to handle X, call Vi to handle Y”
- Set trigger conditions
- Use escalation tool to connect
Advanced play, learn after Month 1.
Nicknames are customizable
Rose / Ada / Vi are realvco defaults. You can change them in Settings sub-tab to:
- “Customer Service Mei”, “Marketing Joe”, “Analyst Zhang” (names for customers)
- “My AI Assistant”, “Home AI” (personal preference)
After changing, Telegram bot and chat interface will use the new name.
Related
- What is realvco — Overall positioning
- Three interfaces overview — Which screens you’ll use
- AI agent tab overview — What you see after clicking into an agent
- Meet the 4 persona files — How to customize persona