AI coding tools
Step by step, connect Continue, Cline, Roo Code, Cursor, Claude Code, Codex and opencode to Helipod AI Gateway models.
Coding assistants that let you set an OpenAI or Anthropic base URL can run on Helipod models. You pay per token from your AI balance, in Rupiah, and use the same kind of key for every tool. This page walks through the most common tools; the full reference, including Aider and other tools, is in Coding tools.
Before you start
Top up and create a key
Top up your AI balance and create a key in AI Gateway → API keys, see AI Gateway quickstart. Create one key per tool, so you can see each tool's spend in Usage and revoke one without breaking the others. A Monthly spend limit on the key is a good safety net for agents.
Keep the key in an environment variable
export HELIPOD_AI_KEY=hel_ai_your_keyPick a model
Open Models & pricing and copy a model id exactly as listed. For agentic coding, start with a larger model (glm-5.3, kimi-k3, deepseek-v4-pro) and use a fast one (deepseek-v4.1-flash, glm-5.3-flash) for background tasks. Any model in Models & pricing works.
| Tool | Format | Base URL |
|---|---|---|
| Continue, Cline, Roo Code, Cursor, opencode | OpenAI chat completions | https://ai.helipod.io/v1 |
| Claude Code | Anthropic | https://ai.helipod.io |
| Codex | Responses API | https://ai.helipod.io/v1 |
Continue (VS Code and JetBrains)
Open the config
In the Continue panel, open the settings and choose to edit the config file, or open ~/.continue/config.yaml directly.
Add a Helipod model
models:
- name: GLM-5.3 (Helipod)
provider: openai
model: glm-5.3
apiBase: https://ai.helipod.io/v1
apiKey: ${{ secrets.HELIPOD_AI_KEY }}
roles: [chat, edit, apply]Use provider: openai with the apiBase above. Add one entry per model you want to switch between.
Chat
Pick GLM-5.3 (Helipod) in the model list of the Continue panel and send a message. Streaming answers work as usual.
Cline and Roo Code
Open the provider settings
Click the settings icon in the extension's panel.
Fill in the provider
| Setting | Value |
|---|---|
| API Provider | OpenAI Compatible |
| Base URL | https://ai.helipod.io/v1 |
| API Key | your hel_ai_... key |
| Model ID | for example glm-5.3 |
Set the model limits
Set the context window and output limit to the model's values from Models & pricing, so the extension does not send more than the model accepts. Then save and start a task.
Kilo Code uses the same settings.
Cursor
In Settings → Models, enter your hel_ai_... key as the OpenAI API key, turn on Override OpenAI Base URL with https://ai.helipod.io/v1, and add the model ids you want (for example glm-5.3). Which Cursor features use custom models depends on your Cursor plan.
Claude Code
Claude Code speaks the Anthropic format. Set these variables, then run claude:
export ANTHROPIC_BASE_URL=https://ai.helipod.io
export ANTHROPIC_AUTH_TOKEN=$HELIPOD_AI_KEY
export ANTHROPIC_MODEL=glm-5.3
export ANTHROPIC_DEFAULT_HAIKU_MODEL=deepseek-v4.1-flash # background tasks
claudeNote the base URL has no /v1 here. Put the lines in your shell profile to keep them. See Anthropic compatible for details.
Codex
Codex uses the Responses API. Add Helipod as a provider:
model = "glm-5.3"
model_provider = "helipod"
[model_providers.helipod]
name = "Helipod AI Gateway"
base_url = "https://ai.helipod.io/v1"
env_key = "HELIPOD_AI_KEY"
wire_api = "responses"Then run codex as usual. A notice that Codex has no metadata for the model can be ignored.
opencode
Add a provider to ~/.config/opencode/opencode.json with "npm": "@ai-sdk/openai-compatible", baseURL set to https://ai.helipod.io/v1 and apiKey set to {env:HELIPOD_AI_KEY}, and list the models you want. The full file is in Coding tools. Choose the model with /models.
Troubleshooting
| Error | Cause | Fix |
|---|---|---|
401 | The key is missing, wrong, revoked or expired. | Copy the key again from API keys (no spaces), or create a new one. |
402 | Your AI balance is empty. | Top up under Overview, and turn on Auto top-up. |
403 | The key reached its monthly spend limit. | Wait for next month, raise the limit with a new key, or use another key. |
404 model not found | The model id is wrong, switched off, or not allowed for this key. | Use the model id exactly as listed in Models & pricing, and check the key's allowed models. |
404 not found | The base URL is wrong. | Use https://ai.helipod.io/v1 for OpenAI-style tools and https://ai.helipod.io for Claude Code. |
429 | Too many requests at once. | Slow down; the tool usually retries by itself. |
Long agent sessions cost tokens fast, because each turn resends the conversation. Check Usage now and then, and use a fast model for background tasks.