Moonshot

Moonshot AI serves the Kimi model family, with long context windows and separate thinking variants for reasoning-heavy work.

Website: platform.moonshot.ai (international) | platform.moonshot.cn (mainland China)

Get an API key

International
Mainland China

Go to platform.moonshot.ai and sign up or sign in.
Open the API key management page in the console.
Create a new API key.
Copy the key immediately and store it securely.

Configure Dline

See Model configuration for the general flow.

Click the Settings button in the Dline panel title bar.
Set Provider to Moonshot.
Under API Version, keep v1, the current version of the Moonshot API.
Paste the key into the Moonshot API Key field.
Select a model under Model.

API address

Dline sends requests to https://api.moonshot.ai/v1, the international host of the Moonshot API. The mainland China platform at platform.moonshot.cn runs its own deployment with its own keys, so use a key issued by the console that matches the host above.

Models

Model	Context window	Notes
`kimi-k2.5`	262,144	Accepts images and caches prompts.
`kimi-k2-0905-preview`	262,144	Default model.
`kimi-k2-0711-preview`	131,072
`kimi-k2-turbo-preview`	262,144	Faster variant of `kimi-k2-0905-preview`.
`kimi-k2-thinking`	262,144	Streams reasoning content that Dline shows as thinking.
`kimi-k2-thinking-turbo`	262,144	Faster variant of `kimi-k2-thinking`.

Only kimi-k2.5 caches prompts, so the usage panel shows cache reads and writes for that model and reports the other Kimi models with plain input tokens.

The Moonshot card has no thinking control. A thinking model always reasons before it answers, and the model decides on its own.

Troubleshooting

Problem	What to do
`401` on every request	The key was not accepted by the API host above. Confirm the key is active in the console you issued it from, and that the console matches `https://api.moonshot.ai/v1`.
`404` on a model	The key has no access to that model ID. Pick another model, or check which models your account can call in the console.
`429`	The account hit its rate limit or ran out of balance. Check the console and try again later.
A long prompt is refused	The history exceeds the context window of the selected model. Start a new task, compact the context, or pick a model with a larger window.