Kimi Code CLI with Nbility

Kimi Code CLI accepts custom OpenAI-compatible providers. This configures CLI inference, not the Kimi web account or subscription.

Configure a provider and model aliasCopy link

Use /provider and Add New Platform for interactive credential setup. For manual configuration, edit ~/.kimi/config.toml:

default_model = "nbility-chat"

[providers.nbility]
type = "openai_legacy"
base_url = "https://api.nbility.ai/v1"
api_key = "YOUR_API_KEY"

[models.nbility-chat]
provider = "nbility"
model = "YOUR_MODEL_ID"
max_context_size = 8192

Replace the key, model ID, and context limit. 8192 is illustrative. This file contains credentials; keep it in your user directory with restricted access and out of Git.

default_model refers to the local alias nbility-chat; model is the ID sent to the gateway. Quote TOML names containing dots, for example [models."my.model"].

Choose the protocolCopy link

• Chat Completions uses openai_legacy.
• Responses uses openai_responses and a compatible model.
• Do not reuse the built-in Moonshot provider’s assumptions for this gateway.

VerifyCopy link

Start kimi, confirm the alias with /model, send a short message, and inspect Usage logs. Check saved keys for authentication failures and context limits for early compaction. Never exceed the actual model window.

ReferencesCopy link

Adapted and rewritten for Nbility from aicentos’s Kimi Code CLI guide, licensed under CC BY 4.0. Its account, plan, and proprietary service instructions do not apply here. See the client’s official documentation.