Qoder Custom Models: Claude & GPT

Qoder is Alibaba's agentic IDE; add Code By AI as an OpenAI-compatible provider to run Claude and GPT models with one Code By AI key.

Qoder supports bring-your-own-key models through its Models settings. The custom-provider form accepts an OpenAI-compatible base URL, so Code By AI can serve both model families behind one entry: GPT IDs go to the Codex channel, and claude-* IDs are forwarded to the Claude channel and billed at Claude prices.

One command?

setup.sh does not configure Qoder. Its BYOK providers and credentials are managed by the Qoder account catalog, and the Qoder CLI documentation explicitly says not to write them into settings.json. Use the UI steps below.

1. Create a Code By AI API key

Sign in to Code By AI, open API Keys, and create a key. It looks like sk-proj-....

2. Add a custom provider

1. Open Qoder Settings.
2. Select Models in the left navigation.
3. Click Add model.
4. Under Provider, choose Custom → OpenAI Compatible, then Chat Completions API.
5. Enter the Base URL https://www.codebyai.net/api/agent/v1.
6. Paste the Code By AI API key.
7. Add the model ID gpt-5.6-terra.
8. Optionally set a Display name.
9. Configure the context window and capabilities Qoder shows for the model. Leave the defaults when unsure. Enable Vision or Thinking only for models that actually support them.
10. Click Validate and add model.

Qoder expects a base URL, not a full endpoint. Paste /api/agent/v1 exactly; Qoder adds the Chat Completions path itself.

3. Add Claude models to the same provider

Click Add Model ID and enter claude-sonnet-5. The same base URL serves both families, so one provider entry is enough for Claude and GPT. Add every ID you plan to use from the Model Plaza.

If you choose Responses API instead of Chat Completions API, keep the same base URL: the gateway also routes /api/agent/v1/responses. Chat Completions is the safer default for third-party OpenAI-compatible clients because it is the format almost every provider implements.

4. Select the model

Open a chat, choose the new model in the model picker, and send a short request. A normal reply confirms the key, base URL and model ID.

Billing

Custom-model traffic is billed by Code By AI, not by Qoder Credits. Qoder's Repo Wiki keeps using its own fixed model and still consumes Credits; adding a custom model does not change that.

Troubleshooting

• Validation fails — check the Base URL has no /chat/completions suffix, the key has no leading or trailing spaces, and the model ID matches the Model Plaza exactly.
• 404 — the Base URL is a full endpoint or contains a duplicated /v1. Use https://www.codebyai.net/api/agent/v1.
• 401 — the key is invalid, deleted or expired.
• Insufficient balance — top up on the Wallet page.
• Claude model answers with GPT prices or behaviour — confirm the model ID starts with claude-; the gateway chooses the channel from that prefix.

Use one route per model. Adding the same model ID under both Chat Completions and Responses makes failures harder to diagnose and does not add capability.