AI Coding
AI Coding lets you connect popular AI coding tools — Claude Code, OpenAI SDK, IDE extensions — directly to models operated by GreenNode, using cloud models without managing API keys from external providers.
Architecture
Requests from your tool are redirected to a GreenNode endpoint. The endpoint exposes two protocols in parallel to support all existing clients:

A single key works for both protocols — but the key must match the host of the service type that issued it.
Two service types — settle this before configuring anything
PAYG
Pay per token actually used
API Key from API Keys
maas-llm-aiplatform-hcm.api.vngcloud.vn
Token Plan
Prepaid 30-day package with a fixed token quota
subscription-key from Plan Detail → Subscription keys tab
tokenplan.api.greennode.ai
Base URL by service type and client standard
Service type
Anthropic-standard clients (Claude Code, Anthropic SDK)
OpenAI-standard clients (OpenAI SDK, LiteLLM, Cursor, Continue.dev, Codex, OpenCode)
PAYG
https://maas-llm-aiplatform-hcm.api.vngcloud.vn
https://maas-llm-aiplatform-hcm.api.vngcloud.vn/v1
Token Plan
https://tokenplan.api.greennode.ai
https://tokenplan.api.greennode.ai/v1
Picking the wrong cell in this table is the most common failure: the wrong /v1 suffix gives 404 Not Found; the wrong host for your key type gives 401 Unauthorized even while the key is still valid. Every tool guide has separate tabs per service type — pick your tab and copy it as-is.
Supported Tools
Claude Code
Claude Code CLI supports overriding ANTHROPIC_BASE_URL — pointing to a GreenNode endpoint instead of Anthropic directly. All sessions, tool calls, and sub-agents route through the GreenNode endpoint, with usage visible in AI Platform Console.
OpenAI-compatible clients
Any tool that allows setting a custom base_url in OpenAI SDK format works out of the box — OpenAI Python/Node.js SDK, LiteLLM, Cursor, Continue.dev, and other IDE extensions. Change the base URL and key; no logic changes needed.
Billing
Pricing
Credit-token, 1 credit = 1 VND
Fixed prepaid 30-day package
Quota
Prepaid: credits deducted every 5-minute collection cycle; model auto-disabled when credits run out. Postpaid: usage recorded as debt with no quota limit
Fixed token/request quota per model, shared across all subscription-keys in the plan
Getting Started
Prepare a key and Base URL, and choose a model
Use a GUI tool
Use a CLI tool
Configure an SDK / IDE extension on the OpenAI standard
Attach an MCP server to an agent
Buy and use a Token Plan package
Get a PAYG API key
Last updated

