Pre-installed API Credit & OpenRouter Setup
Checkout step 2 asks whether you want to add a pre-installed API credit ($0 / $5 / $10). This page explains what it actually is, why it’s designed this way, and how to use it to bootstrap your server properly.
TL;DR: The pre-installed API credit is an OpenRouter LLM allowance we buy and install for you ahead of time, so your AI agents can already talk to you the moment your server boots — and then help you finish all the real configuration. It’s a bootstrap helper, not a long-term LLM subscription.
Why we designed this
If you install OpenClaw or Hermes yourself, the first hurdle looks like this:
- You sign up at OpenAI / Anthropic / Google, request API keys, top up
- You edit OpenClaw or Hermes config files to inject the keys
- Only then can the AI start talking to you
- Then you go set up Telegram / WhatsApp / LINE and other messaging platforms
- The whole bootstrap process, the AI is mute and can’t help you
realvco does it differently:
- At checkout, pick $0 / $5 / $10 pre-installed credit
- The moment your server is built, the AI already has working credit
- You sign in to admin panel and the AI is right there waiting to chat
- You tell Rose “help me set up your own production OpenAI API key” → AI walks you through signup, key creation, and config edits
- Once your real key is connected, AI runs on your own key and the pre-installed credit’s job is mostly done
The goal: kill the “mute AI during setup” pain point.
Three-way decision table
| Credit | One-time | What it can do during the 90-day window |
|---|---|---|
| $0 | Skip | You already have your own OpenRouter / OpenAI / Anthropic key. AI won’t respond when the server boots — set the keys up yourself before your lobsters can start working |
| $5 (default) | One-time $5 | Enough for AI to install production LLM API keys for all three lobster agents, possibly enough for Telegram / WhatsApp setup too |
| $10 | One-time $10 | Definitely enough for both production LLM keys and messaging platforms, plus extra room for AI to walk you through more setup details |
Key rules — different from what you might expect:
- One-time purchase: $5 means $5 total, not $5/month, not monthly resets
- 90-day window: the pre-installed key expires after 90 days regardless. Unused balance is forfeited, no refunds, no rollover
- Available only once: this is an add-on service that can only be purchased together with your first subscription to a new host. You can’t buy another later
What you’re buying
When you pick $5 or $10, at the moment checkout completes realvco does this:
- Generates an OpenRouter sub-key for you
- Sets the cap = the amount you bought
- Sets a 90-day expiry
- Saves it to your customer record
- Injects it into all 3 AI containers (oc-1 Rose / hm-2 Ada / hm-3 Vi share the same key) when your server boots
- Pre-configures Rose (OpenClaw) with 4 default models so you can start chatting immediately
The full default model list + how to switch with quick aliases (gm / ds / hk / gf): see Pre-installed Model List.
Ada / Vi (Hermes-Agent) get the same key but no default model list — first time using Ada / Vi, pick a model from the Settings sub-tab in admin-panel. See Hermes TUI via the web terminal.
Recommended flow ($5 / $10 customers)
The proper way to use the pre-installed credit is to let AI relay-install everything else for you:
Step 1: Sign in and talk to Rose first
Once your server is up and you’ve signed in to admin panel, Rose can already chat. Open with something like:
“Rose, help me set up your own production OpenAI API key — walk me through it.”
Rose will guide you through signup, key creation, and writing the key into her own config.
Step 2: Use the now-self-sufficient Rose to relay-install the others
Once Rose has her real key, she’s running on your key, not the pre-installed credit. Use Rose to install Ada and Vi:
“Rose, please set up Ada with her own production API key too."
"Rose, while you’re at it, configure Vi as well.”
Step 3: Connect messaging platforms
With all 3 agents on production keys, ask the AI to wire up Telegram / WhatsApp / LINE / Microsoft Teams etc.:
“Ada, connect yourself to Telegram — I want to chat with you on my phone.”
Step 4: Pre-installed credit retires gracefully
By this point, all 3 agents run on your own production keys. The pre-installed credit may have a small balance left as a safety net, or may be nearly depleted — either way, you’re self-sufficient.
Why is $5 usually enough?
During bootstrap, the AI is mostly doing short “guide you through signup + edit config” conversations. $5 typically covers production API installation for all 3 agents plus the major messaging platforms. What if you run out before finishing? No problem — see the next section.
What if you run out
When the pre-installed credit runs out, all 3 agents go “running but silent”. But this isn’t necessarily bad — by that point you’ve usually got at least one agent on a real key already:
| Your situation | What to do |
|---|---|
| At least 1 agent already on a production key | Use that agent to install the others. The pre-installed key going dead doesn’t affect agents running on your own keys |
| None installed yet (rare edge case) | Sign up at openrouter.ai for your own key (see next section). The pre-installed credit is a one-time add-on at first subscription — you can’t buy another for the same host afterwards. For unusual situations, contact realvco support to discuss exceptional assistance |
The 90-day expiry follows the same logic: the pre-installed key dies, but agents on your own keys are unaffected.
Bringing your own OpenRouter key (continuation path)
If your pre-installed credit expires and you want to keep using OpenRouter rather than switching to OpenAI / Anthropic direct, the path is simple:
- Sign up at openrouter.ai
- Create a key at openrouter.ai/keys
- Add credit (pre-paid, $10 minimum)
- SSH into your host, edit
/opt/openclaw/.tokens→ setOPENROUTER_API_KEY - Restart the 3 containers
Rose’s default model aliases carry over — your own new key calls the same aliases, no reconfig needed. See the full alias list at Pre-installed Model List.
For OpenRouter account / billing / security details, see the OpenRouter docs.
Bringing your own provider key (OpenAI / Anthropic direct)
You can also bypass OpenRouter and connect directly to OpenAI or Anthropic. This is what large-volume production customers usually pick:
| What you want | Where to edit | See |
|---|---|---|
| OpenClaw (Rose) → OpenAI / Anthropic direct | Container’s openclaw.json models section | docs.openclaw.ai Persona / Models |
| Hermes (Ada / Vi) → other provider | Hermes .env HERMES_LLM_* | Hermes-Agent upstream GitHub docs |
Why does pre-install use OpenRouter instead of OpenAI / Anthropic direct?
The bootstrap phase needs “fast start + try multiple models”. OpenRouter’s “one key, 100+ models” is the best fit. Once you’ve used the AI to walk through each provider’s setup yourself, you can decide whether to stick with OpenRouter long-term or go single-provider direct.
FAQ
Q: Will my conversation history disappear when the pre-installed credit expires?
No. Conversations, persona settings, and automatic backups stay intact. Only the conversations running on the pre-installed key go silent — agents on your own production key are unaffected.
Q: Is $5 really enough?
Depends on how much you lean on AI for setup hand-holding. Most customers can install all 3 agents’ production LLM API keys plus the major messaging platforms on $5. If you want the AI to explain things multiple times or use heavier reasoning models (Haiku 4.5 instead of GPT-5.4 Mini), $10 may be safer.
After bootstrap you switch to your own production key anyway — pre-installed credit was never meant for the long haul.
Q: Why is it a one-time add-on at first subscription with a 90-day expiry?
Because this isn’t an LLM subscription, it’s a bootstrap helper. Allowing endless top-ups would tempt customers to use it as their main LLM — but OpenRouter takes a markup and you miss enterprise discounts from going direct. Limiting it to a one-time add-on at first subscription, with a hard 90-day expiry, is a forcing function to switch to a real production setup.
Q: Can I run a customer-support bot directly on the pre-installed key?
Technically yes, in practice no:
- The 90-day expiry will kill your support bot mid-conversation one day
- $5 / $10 won’t sustain commercial volume
- The system isn’t designed for production use, no extension path
After bootstrap, hook each of the 3 agents to your own production key (OpenAI / Anthropic / your own OpenRouter account).
Q: Does each AI agent get its own key?
No — all 3 containers (Rose / Ada / Vi) share the same pre-installed sub-key, so usage pools together. After you switch to production keys, each agent can run on a different provider (e.g. Rose on OpenAI, Ada on Anthropic, Vi on Google), each billed separately.
Q: Does upgrading / downgrading my server plan reduce my pre-installed credit?
No. The pre-installed credit is independent and tied to its own purchase, not the host plan.
Q: I have my own OpenAI API key (or a ChatGPT subscription). Can I use it?
Your own OpenAI API key: yes. SSH in and edit OpenClaw’s openclaw.json to point models at OpenAI directly instead of openrouter/openai/gpt-.... See OpenAI docs and docs.openclaw.ai.
If you pick $0 + bring your own OpenAI API key, you’re skipping OpenRouter entirely — that’s a perfectly valid advanced path.
⚠️ OpenAI subscription ≠ OpenAI API key
These are commonly confused:
- ChatGPT Plus / Pro / Team / Enterprise subscription: lets you chat with ChatGPT on chatgpt.com / mobile app
- OpenAI API key: a credential for programs (like your AI agent) to call GPT models
They are separate products, billed separately. Your ChatGPT Plus subscription cannot be used to power an AI agent. To run OpenAI models in your AI agent, sign up at platform.openai.com, generate an API key, and add credit — this is unrelated to your ChatGPT subscription.