Skip to main content

Adding a custom AI provider to Kari

Copyright © 2026GotoAI Inc.Based on the goose project

Even for an AI provider that Kari does not ship with, you can add it as a custom provider as long as its API is OpenAI-compatible, Anthropic-compatible or Ollama-compatible. A local LLM running in your office or on your own PC is registered the same way.

  1. On the Configure your AI Provider screen, click Add a custom provider.

The Configure your AI Provider screen

Screenshot shown in the original Japanese.

  1. A screen appears asking how you want to configure it. To pick a known provider and have the settings filled in for you, click Start from a provider template. To enter everything yourself, click Configure manually.

  2. Enter the provider's details. Fields marked with * are required. This guide uses OumiGo, which manages a GPU fleet, as the example backend.

Adding a custom provider

Screenshot shown in the original Japanese.

  • Provider type: choose OpenAI-compatible, Anthropic-compatible or Ollama-compatible to match the shape of the API.
  • Display name: the name shown in Kari's interface.
  • API URL: the provider's endpoint. For a local LLM this is an address such as http://192.168.0.100:7012. For OumiGo, give the Manager's host and router port.
  • API base path (optional): fill this in only to override the default API path. Leave it blank to use the provider's default.
  • Authentication: for providers that need an API key, turn on This provider requires an API key and enter the key. Local LLMs such as Ollama usually do not need one.
  • Available models (comma-separated): enter the model names to use, separated by commas. With OumiGo, the LLM is configured in advance on the Manager node, so any model name will do — the model name is updated as needed during an OumiGo chat session.

  1. Scroll down and you can also set per-model capabilities (Tool calling, Reasoning, Attachments), Context length (tokens), This provider supports streaming responses and Custom headers. Set these as needed.

  2. When you are done, click Create provider to save. The provider you added can then be selected from the list on the Configure your AI Provider screen.

  3. Select the custom provider and model you added, and start working in the chat session.

Starting a chat on a custom provider

Screenshot shown in the original Japanese.

Last updated on