Skip to main content
Kilo Code connects directly to LithosAI through its OpenAI Compatible provider. You do not need a local proxy or protocol adapter.
These steps use the Kilo Code custom-provider interface in VS Code. Kilo uses the same provider configuration across its editor and CLI surfaces.

​
Before you begin

Install the Kilo Code extension, create a key on the API Keys page, and confirm the model ID you want to use on the Models page.
1

Open the provider settings

Open Kilo Code in VS Code, select the gear icon, and go to Providers. Scroll to the bottom of the provider list and select Custom provider.
2

Configure LithosAI

Enter the following values:After you enter the base URL and key, select moonshotai/Kimi-K3 from the fetched model list. If it does not appear, add the model ID manually.
If Kilo does not recognize the model, set its context and output limits (limit.context and limit.output in kilo.jsonc) from the Models page. Without them, automatic context compaction may be disabled.
3

Save and select the model

Select Submit, then choose lithosai/moonshotai/Kimi-K3 in the model picker. Replace the model ID if you selected a different model in the LithosAI console.
{env:...} references resolve only from the trusted global kilo.jsonc or a managed configuration, not from a project-level file.

​
Configure Kilo CLI

If you use the Kilo CLI, install it and add an OpenAI-compatible provider to your trusted global configuration.
~/.config/kilo/kilo.jsonc
Keep this configuration in ~/.config/kilo. Kilo does not resolve {env:...} credentials from an untrusted project-level kilo.jsonc file.
Start Kilo from the project you want it to inspect.

​
Verify the connection

Open a project that contains a README.md, then ask Kilo Code:
Kilo Code should read the local file and return its heading. This verifies both the model response and a basic local tool call.

​
Troubleshooting

Issues common to every agent are listed under Common issues.

​
Resources