A-Z Guide

Who this is for: individuals buying for themselves (B2C), and whoever holds the company's Root/Admin account and needs to give AI access to non-technical departments (B2B).

---

0. Understand it in 2 minutes

For an AI tool (Cursor, OpenCode, Codex, Claude Desktop…) to run on GreenNode models, you only need 2 pieces of information:

What you need	What it is	Analogy
Base URL	The address of the AI server	The street address of the "brain"
API key (called a subscription-key here)	The key that gets you in	The door key

Token Plan is how you get those two on a 30-day prepaid basis: you buy one plan, get a fixed token allowance up front, and know your monthly cost in advance — instead of paying per token actually consumed (PAYG).

Buy a plan  →  Plan auto-creates a subscription-key  →  Copy the key + Base URL
            →  Paste into your tool  →  Run  →  Watch usage in the portal

Want the deeper picture — how Token Plan differs from PAYG and how it's wired together → see Token Plan (overview).

---

1. Prerequisites — check these before you start

Tick all 3 rows before moving to step 2:

#	You need	How to check
1	A GreenNode account that can sign in to AI Platform	Open aiplatform.console.greennode.ai
2	The Root or Admin role	Only these 2 roles can buy plans. See Roles and Permissions
3	Enough Credits in the account to cover the plan price	The balance is shown right on the checkout screen (Balance). 1 credit = 1 VND

If you are a Developer/Viewer, you cannot buy a plan — ask a Root/Admin to buy one and issue you a subscription-key (see Section 6).

Not enough credit? Top up in the AI Platform Console before buying — the system deducts credit the moment you confirm, and does not allow buying on credit.

---

2. Pick a plan

Portal path: AI Platform → sidebar → Token Plan → Packages

1. The Plan Types appear as cards. Each card shows the price, duration, included models, and max subscription-key count at a glance. Example, the Token Plan Alpha card: 1,080,000 VND / 30 days — GLM 5.2 model — up to 5 keys.

2. Click a card to open Package Detail and read it carefully before spending:

The Packages list — one card per Plan Type

Look at	To learn
Max keys	How many people/tools you can issue keys to
Duration	The plan cycle (usually 30 days)
Tokens / Requests per cycle	The allowance per cycle — tracked independently per model, with no transfer between them
Included Models	The plan's model table (Model, Hosting mode, Status, Token quota, Request quota). Click the model name to see its Model code — note it down, you'll need to enter it in your tool shortly
Subscription Endpoint	https://tokenplan.api.greennode.ai — this is the Base URL you will use. OpenAI-standard tools append /v1, Anthropic-standard tools do not — see Section 5
What happens when activated	What happens right after activation

Package Detail — max keys, duration, included models and the Subscription Endpoint

Once purchased, a plan cannot be swapped for a different Plan Type and cannot be returned on request. If you must stop mid-cycle, use Delete — the unused allowance is refunded pro rata to Credits. Think it through at this step.

Which size should you pick? A simple rule for non-technical teams: one person using a coding assistant steadily all day consumes a fair number of tokens — if you have no data yet, buy the smallest plan for 1 cycle to measure, check the real numbers in Section 7, and only then buy bigger.

📎 Full detail on the Packages screen: Buy a Token Plan

---

3. Buy the plan

1. Click Buy Now on the card (or Buy package inside Package Detail).

2. The Confirm & checkout screen appears — fill it in and double-check:

Field	Example	Notes
Plan name (required)	ENG-GLM-1	Only a-z A-Z 0-9 _ - ., max 50 characters. Name it after the department/project so filtering usage later is easy
Plan price / Duration / VAT / Total	1,080,000 VND / 30 days / Included / 1,080,000 VND	Computed from the Plan Type
Auto-renew	ON (default)	Renews at the end of each cycle — can be turned off after purchase
Payment method	Credits	1 credit = 1 VND
Balance	Your current balance	Check that it covers the total

1. Click Confirm & Pay.

Confirming immediately deducts the amount on the Total row. Provisioning runs in the background — the gateway endpoint and subscription-key only appear in full on Plan Detail once provisioning finishes. Wait a few seconds and refresh if you don't see them yet.

Result: the plan appears in My Token Plans with status ACTIVE, Auto-renew ON, and the system auto-creates 1 key named default-key that works right away — no need to create a key manually to get started.

📎 Detail: Buy a Token Plan

---

4. Get the API key + Base URL

Path: API Key → Token Plan → My Token Plans → click the plan name you just bought

The Plan Detail page has 2 tabs — you need information from both:

The Models tab → grab 2 things:

• The Gateway base URL shared by the whole plan — pick the form that matches your tool's standard:

  • OpenAI-standard tools (Cursor, OpenCode, Codex…): https://tokenplan.api.greennode.ai/v1

  • Anthropic-standard tools (Claude Code, Claude Desktop): https://tokenplan.api.greennode.ai — no /v1

• The Model code of the model you want (e.g. glm-5.2) — copy it exactly, one wrong character breaks the call

Plan Detail — Models tab: the gateway base URL and each model's code

The Subscription keys tab → grab the key:

• Copy default-key (or a key you created yourself in Section 6)

Plan Detail — Subscription keys tab

Jot these 3 values down:

Information	Value
Base URL	https://tokenplan.api.greennode.ai/v1 (OpenAI standard) or https://tokenplan.api.greennode.ai (Anthropic standard, no /v1)
API key	the subscription-key you just copied
Model	the model code from the Models tab (e.g. glm-5.2)

A subscription-key is a secret — anyone holding it can call models and burn the plan's allowance. Don't paste it into group chats, don't commit it to Git. Issue a separate key per person (see Section 6) so a leak only costs you one key.

---

5. Set it up in the tool running your agent

Token Plan uses one shared host for every tool — the two standards differ only in the /v1 suffix and in the environment variable names:

Tool standard	Base URL to enter
OpenAI (Cursor, Continue.dev, OpenCode, Codex, LiteLLM, OpenAI SDK…)	https://tokenplan.api.greennode.ai/v1 — with /v1
Anthropic (Claude Code, Claude Desktop, Anthropic SDK)	https://tokenplan.api.greennode.ai — without /v1

Nothing else changes — you still only need the 3 values from Section 4.

5.1 OpenAI-standard tools with built-in config fields (Cursor, Continue.dev, and similar)

In the tool's Settings, fill in:

Field	Value
Base URL	https://tokenplan.api.greennode.ai/v1
API Key	<your subscription-key>
Model	<model code> — e.g. glm-5.2

5.2 OpenAI-standard tools that read environment variables (CLI)

macOS / Linux / WSL

export OPENAI_BASE_URL="https://tokenplan.api.greennode.ai/v1"
export OPENAI_API_KEY="<your subscription-key>"

Add those 2 lines to the end of ~/.zshrc (macOS) or ~/.bashrc (Linux/WSL) so you don't retype them every time, then run source ~/.zshrc.

Windows PowerShell

$env:OPENAI_BASE_URL = "https://tokenplan.api.greennode.ai/v1"
$env:OPENAI_API_KEY  = "<your subscription-key>"

To persist it for your account (run once, then reopen PowerShell):

[Environment]::SetEnvironmentVariable("OPENAI_BASE_URL", "https://tokenplan.api.greennode.ai/v1", "User")
[Environment]::SetEnvironmentVariable("OPENAI_API_KEY", "<your subscription-key>", "User")

5.3 Calling it from code (OpenAI SDK)

from openai import OpenAI

client = OpenAI(
    base_url="https://tokenplan.api.greennode.ai/v1",
    api_key="<your subscription-key>",
)

response = client.chat.completions.create(
    model="glm-5.2",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

5.4 Test the connection (30 seconds)

Paste this whole block into a terminal, replacing the 2 values in <>:

curl https://tokenplan.api.greennode.ai/v1/chat/completions \
  -H "Authorization: Bearer <your subscription-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<model code>",
    "messages": [{"role": "user", "content": "ping"}]
  }'

✅ It works when: the JSON response contains a choices[0].message.content field.

5.5 Common errors

Symptom	Cause	Fix
401 Unauthorized	Wrong key, key Disabled, or characters missing from the copy	Re-copy the key from the Subscription keys tab, check Status = ACTIVE
403 Forbidden	Calling a model not included in the plan	Only call models listed in that plan's Models tab
402 Payment Required	The plan has expired or was deleted	Buy again or turn Auto-renew back on
404 Not Found	The Base URL doesn't match your tool's standard	OpenAI-standard tools: the Base URL must end with /v1. Anthropic-standard tools (Claude Code): the Base URL must not have /v1
Requests run out of tokens mid-cycle	That model's token pool hit 0	Wait for the new cycle, buy another plan, or switch to a PAYG API Key for now
Response parse error	The tool appends /v1 to the base URL itself	Try dropping /v1 if the tool handles it for you

5.6 Anthropic-standard tools (Claude Code, Claude Desktop)

These tools speak the Anthropic protocol rather than OpenAI, so the Base URL is the same host without /v1, and the environment variable names differ too:

macOS / Linux / WSL

export ANTHROPIC_BASE_URL="https://tokenplan.api.greennode.ai"
export ANTHROPIC_AUTH_TOKEN="<your subscription-key>"

Windows PowerShell

$env:ANTHROPIC_BASE_URL   = "https://tokenplan.api.greennode.ai"
$env:ANTHROPIC_AUTH_TOKEN = "<your subscription-key>"

Full walkthrough: Claude Code.

📎 Detailed configuration for specific tools (LiteLLM, Cursor, Continue.dev, Node.js SDK…): Connect OpenAI-compatible Clients to GreenNode MaaS — the steps stay the same, just point the Base URL and key at your Token Plan values from Section 4.

📎 Supported tools and how to pick the right one: AI Coding · Getting Started with AI Coding · CLI Tools

---

6. Manage: issue a key to each member

This section is for whoever holds Root/Admin and needs to share a plan across a team.

6.1 View your purchased plans & each plan's detail

Path: API Key → Token Plan → My Token Plans

The list shows: plan name, Plan Type, expiry date (Expires), and Auto-renew status.

The My Token Plans list

Status	Meaning
ACTIVE	The plan is running
EXPIRED	The plan expired without renewal — every key in it stops working

Click a plan name to open Plan Detail:

• General information: Package, Price, Purchase date, Expiry date, Cycle, keys in use (e.g. 1/5 in use), Auto-renew, and the Tokens/cycle and Requests/cycle allowances

• Models tab: Gateway base URL + the model table (Model, Model code, Enabled types, Endpoint) — everything you need to call the API

• Subscription keys tab: the keys in this plan (Name, Status, Key, Created at)

• Actions, top right: Buy again (buy another plan of the same type) and Delete (cancel the plan)

Buying several plans in parallel gives you several independent plan instances — quota is never pooled, even across the same Plan Type. To split budget by department, buy one plan per department and name each plan after it.

6.2 Issue a subscription-key to each member

The rule: one key per person / per tool. They share the plan's allowance, but separate keys let you revoke per person and filter usage per person.

Step 1 — Plan before you create. The plan has a Max keys limit (say 5), so list who needs what first:

Key name	Issued to	Used for
default-key	(created by the system)	Keep for testing, don't hand it out
mkt-an	An — Marketing	Cursor on a personal machine
sale-binh	Bình — Sales	Codex Desktop
ops-agent-prod	(not a person)	An automated agent running in the background

Name them {department}-{name} or {system}-{environment} — the name alone should tell you which one to revoke.

Step 2 — Create the key.

1. Go to Plan Detail → Subscription keys tab → click + Add key

2. Fill in Key name following your plan table

3. The popup shows what the key inherits from the plan: access to every model in the plan, the plan's shared token/request allowance, and the same gateway endpoint

4. Click Create key

Create subscription key popup

The new key appears with status ACTIVE and automatically shows up on the Access Control page with Key Type = Plan: {plan name} — the same record, no duplicate.

Step 3 — Hand the key to the member. Send each person exactly 3 values:

Base URL : https://tokenplan.api.greennode.ai/v1   # OpenAI-standard tools (Cursor, OpenCode, Codex…)
           https://tokenplan.api.greennode.ai      # Anthropic-standard tools (Claude Code, Claude Desktop)
API key  : <that person's own key>
Model    : <model code>

Include a link to Section 5 so they can set it up themselves.

Send keys over a private channel (1-1, a password manager), never in a shared channel.

Step 4 — Revoke / edit a key. In the Subscription keys tab, click the ⋮ icon on the key's row → choose Enable / Disable / Rename / Delete.

Situation	What to do
Member leaves / changes project	Disable (temporarily) or Delete (permanently)
Suspected key leak	Delete the old key → + Add key to create a new one, then resend
Wrong name	Rename
Out of Max keys and need to add someone	Delete an unused key, or Buy again to buy another plan

Disable and Delete take effect immediately — that person's tool stops being able to call models from the very next request.

6.3 Turn Auto-renew on/off

Click the Auto-renew toggle in the list or in Plan Detail. The Turn off auto-renew? popup spells out:

• The plan will not renew, but stays ACTIVE until its exact expiry date

• On the expiry date, every subscription-key in the plan stops working immediately

• No credit is charged now, and no refund is issued for turning Auto-renew off

• You can turn it back on, or Buy again, anytime before the plan expires

Turn off auto-renew popup

With Auto-renew ON, the system renews before expiry. If credit is short, it retries — and if that still fails, the plan becomes EXPIRED and the whole team loses model access. Keep enough credit on hand before the expiry date.

6.4 Delete/cancel a plan

In Plan Detail click Delete, or in My Token Plans tick the checkbox and click the trash icon (bulk). The popup spells it out: every key is torn down immediately, the unused allowance is refunded pro rata to Credits, and the plan disappears from the list while usage history is retained for auditing.

Delete Token Plan popup

Deleting a plan cannot be undone. Tell the team before you click.

📎 Full detail for Section 6: Manage Token Plan

---

7. Track usage per plan

Path: AI Platform Console → sidebar → Usage & Budget → Usage & Cost → Usage tab

With Token Plan you only look at the Usage tab — not the Cost tab. The plan is prepaid: the cost was fixed at purchase and no extra charge accrues per request. That's why subscription-keys have no per-token cost breakdown like PAYG API Keys do. What matters is how much of the allowance you already paid for has been consumed, not how much money was spent.

7.1 Filtering to the right plan

1. Open the All API Keys dropdown on the filter bar (single select)

2. Select the subscription-key of the plan you want to inspect

3. Set the Time Range to match the plan cycle — use the Absolute tab in the picker panel, set From = purchase date, To = today

Every figure on the Usage tab immediately narrows to that key.

The All API Keys filter is single select — one key at a time. To get a total for a plan with several keys, add the keys up manually, or compare directly against the remaining allowance already shown in Plan Detail.

7.2 Reading the Usage tab

Metric	What it means for a Token Plan
Tokens Consumed	The number that matters most — total Input + Output + Cache. Compare against Tokens/cycle in Plan Detail to see what's left
Total Requests	Compare against Requests/cycle in Plan Detail
Errors	Failed requests — a sudden spike usually means a wrong key, a model outside the plan, or an exhausted allowance
Requests over time	Line chart by hour/day — shows whether usage is steady or bursty
Token Breakdown	Doughnut splitting Cache / Output / Input

The most accurate source for remaining allowance is still the plan's Plan Detail (Tokens/cycle, Requests/cycle). Use the Usage tab to break it down per key — to see who is consuming which share of the shared pool.

7.3 What to do with the numbers

Question	How to answer it
Who is using the most?	Filter each member's key in turn and compare Tokens Consumed
Is the plan about to run dry?	Compare Tokens Consumed against Tokens/cycle in Plan Detail
Should next cycle be a bigger or smaller plan?	Look at total consumption over a full cycle (Time Range = Absolute, purchase date to expiry date)
Is anyone erroring out repeatedly?	Check the Errors card while filtered to that key

📎 Dashboard detail, the other filters, and per-role view permissions: View Usage & Cost

---

8. FAQ

Several people share one plan — how is the quota split? It isn't. Every key in the same plan shares one pool of tokens/requests per model — first call, first served. For a hard split by department, buy a separate plan per department.

What if tokens run out mid-cycle? Calls to that model pause until the new cycle. To keep going right away: buy another plan, or switch to a PAYG API Key for now.

Do subscription-keys show up alongside PAYG API Keys? No. The two are entirely separate and never share a dropdown anywhere in AgentBase. Subscription-keys are also excluded from the Rate Limit page.

Can I call a model outside the plan? No — the gateway blocks it with 403 Forbidden. Only models in the plan's Models tab can be called.

Can I switch to a different Plan Type mid-cycle? No. You have to Delete the old plan (unused allowance refunded pro rata) and buy a new one.

---

What's next

I want to...	Go to
Understand Token Plan vs PAYG and how it's wired	Token Plan (overview)
Detail on the Packages screen & checkout	Buy a Token Plan
Detail on managing plans & subscription-keys	Manage Token Plan
Detailed setup for each AI coding tool	Connect OpenAI-compatible Clients to GreenNode MaaS
Pick the right tool (GUI or CLI)	Getting Started with AI Coding
The usage dashboard (Usage tab)	View Usage & Cost
Manage member permissions in the organization	Roles and Permissions

---

Need help?

• Email: [e-mail]

• Hotline: [phone]

• Support center: helpdesk.greennode.ai

Thank you for using GreenNode services.