Use asAi with Hermes Agent in Malaysia.
Hermes Agent runs on your device and can call an OpenAI-compatible model endpoint. asAi supplies the models and prepaid credits in MYR. This guide shows the exact base URL, key steps, model picks and RM cost checks. Hermes Agent is an open-source project by Nous Research. asAi is not affiliated with Nous Research.
You keep the agent workflow on your computer. asAi handles the model request, checks your prepaid balance, records the usage and exposes the result through the same OpenAI-compatible route. You can change models without creating another account or moving to a monthly plan.
Malaysian developers pay in MYR through FPX or DuitNow, so there is no foreign card charge and no monthly billing to cancel. A single RM5 pack is enough to complete this whole guide, run the verification prompt and explore several models before deciding whether to top up more.
Before you start
- Install Hermes from the official installation guide and finish its setup.
- Create an asAi account and add prepaid credit from RM5 with FPX or DuitNow under Top up.
- Create an API key under API keys. It starts with
asai_sk_. Copy it once and store it privately.
Never share the key in chat, screenshots, public files or support messages. If it is exposed, revoke it immediately and create a replacement. Treat the key like a password because requests made with it spend your balance.
Set aside five minutes and keep your phone ready if your bank requires FPX approval during top-up. Once the balance shows in the dashboard, you will not need to touch payment again until the pack expires after 720 hours.
Connect a custom endpoint
Run hermes model in your terminal. Choose the custom OpenAI-compatible endpoint in the provider picker. Menu wording can vary by Hermes version. Follow the current Hermes model docs if your menu looks different.
Base URL: https://serveras.click/v1
Key: asai_sk_... (pasted privately in hermes model)
Example model: asai/gpt-6-sol (check live catalog before use)
- Enter
https://serveras.click/v1as the base URL. The website isserverasai.com. The API domain isserveras.click. - Paste your asAi key only when Hermes prompts for the credential. Do not commit it to Git, place it in a screenshot or add it to a shared environment file.
- Choose Chat Completions compatibility, then pick a current alias from the model catalog, for example
asai/gpt-6-solif listed. - Confirm the saved provider, base URL and model before you exit the picker. Leave automatic context detection enabled when Hermes offers it.
This endpoint is for OpenAI-compatible Chat Completions. Codex can require its Responses-compatible setup, while Claude Code uses the Messages-compatible guide. Use the dedicated Codex or Claude Code instructions rather than reusing the Hermes settings blindly.
The base URL ends at /v1. Do not append /chat/completions in the Hermes provider field because Hermes adds the request path itself. Do not enter https://serverasai.com either; that host serves the account website, not model traffic.
Pick a model for the task
Start with one model for one task. Run hermes model again when the job changes. Switching inside a long conversation can reset prompt cache, so start a new session when predictable usage matters.
| Task | Model alias | Why pick it |
|---|---|---|
| Complex code and planning | asai/gpt-6-sol | Large context and a higher rate for harder work. |
| Careful analysis and long documents | asai/claude-opus-5 | Strong reasoning for sustained reading and synthesis. |
| High-value code output | asai/deepseek-v4-pro | Lower weighted rate for substantial coding runs. |
| Fast everyday extraction | asai/gpt-6-luna | Lower rate for routine, bounded prompts. |
Always confirm the exact alias in Models on the day you run it. Aliases can retire and new models can appear. The catalog also shows context length, input type and current input, cache and output multipliers.
Use a budget model for repeated extraction, classification or short transformations. Reserve a higher-rate model for architecture, difficult debugging and long planning sessions where output quality matters more than request count. You can keep several saved provider choices and select the one that matches the next job.
What RM5 buys
RM5 pack gives 80,000,000 credits for 720 hours. No subscription. Every available model shares the same balance at its published input, cache and output rates.
- Illustrative example A: 200 input + 650 output tokens on
asai/gpt-6-sol= 34,500 credits, about RM0.00216, about 2,318 similar requests per RM5 pack. - Illustrative example B: 10,000 input + 1,000 output tokens on
asai/claude-opus-5= 225,000 credits, about RM0.01406, about 355 similar requests per RM5 pack. - Illustrative example C: 100 input + 1,500 output tokens on
asai/deepseek-v4-pro= 38,000 credits, about RM0.00238, about 2,105 similar requests per RM5 pack.
Recheck exact multipliers in Billing before you publish cost claims. Actual deduction depends on cache and generated output.
Check the connection
Open a new Hermes session and send a small, non-sensitive prompt. A model reply means the route works. Then open Logs and confirm the model, token counts, timestamp and credits. A paid test consumes credit, so keep the first prompt short and avoid attaching private files.
Check the request ID if you need support. It lets the team locate a specific call without receiving your prompt or secret key. To switch the default model later, run hermes model again, select the saved asAi provider and start a new session.
If something fails
- 401: The key is wrong or revoked. Recheck the provider setup. Revoke any key that appeared in chat, a screenshot or a repository.
- 402: The balance is exhausted or the active pack expired. Check Top up and the account balance.
- Model not found: The alias is wrong or retired. Compare it character by character with the live model list.
- Connection error: The base URL must be exactly
https://serveras.click/v1, not the website domain and not a model-specific URL.
If a request still fails, contact asAi support on Telegram with the time, model ID and request ID. Include the Hermes version and the visible error. Never send the secret key.
Hermes Agent and asAi FAQ
Is asAi affiliated with Hermes Agent?
No. Hermes Agent is by Nous Research. asAi is an independent API provider with an OpenAI-compatible endpoint.
What base URL do I use?
Use https://serveras.click/v1 for Chat Completions. The website serverasai.com is not the API base.
Can RM5 use every model?
Yes. Package size does not gate model access. Availability and account controls still apply.
How long do credits last?
Credits last 720 hours from top-up. There is no auto renewal.
How do I pay from Malaysia?
Pay by FPX or DuitNow through Top up. Packs start from RM5.
Where do I check usage?
Logs shows the model, token counts, timestamp and credits for each request.
Continue with asAi
Read Models before selecting an alias, check the current formula in Billing, or return to the Quickstart for the shortest connection path.
Start with RM5, create a key, connect Hermes tonight.