Get started

Configure an AI provider

Use Kition Cloud or connect OpenAI, Anthropic, DeepSeek, Kimi, or a supported custom endpoint.

Updated

By Kition Docs

Verified against Kition 0.2.0 (7d49036f) on 2026-11-06

Choose how AI should run

  • Kition Cloud — sign in with a Kition Account; no separate API key is required, and usage is deducted from your Kition credits
  • OpenAI — Responses API with an OpenAI API key
  • Anthropic — Messages API with an Anthropic API key
  • DeepSeek — DeepSeek’s Anthropic-compatible Messages API with a DeepSeek API key
  • Kimi — Moonshot’s OpenAI-compatible API with a Kimi API key
  • Custom — your own endpoint, using the OpenAI Responses, OpenAI Chat Completions, or Anthropic Messages API

Open AI Providers

Open Settings → AI Providers. Select a built-in provider or choose Add custom, enter the API key, and choose Save provider.

Then choose Sync now to retrieve the provider’s model list, and pick the Default model used by new agents and chats. You can switch the model per chat at any time.

Built-in defaults

  • OpenAI base URL: https://api.openai.com/v1
  • Anthropic base URL: https://api.anthropic.com
  • DeepSeek base URL: https://api.deepseek.com/anthropic
  • Kimi base URL: https://api.moonshot.cn/v1
  • Every provider has an editable Models endpoint under Advanced
  • Custom providers also set the Base URL, Auth scheme (Bearer, Raw, or X-API-Key), Wire API, and Reasoning effort

Wire compatibility

A custom provider must implement the Wire API you select: OpenAI Responses, OpenAI Chat Completions, or Anthropic Messages.

Heads up

Do not select a wire based only on the provider’s marketing name. Confirm that the endpoint implements the configured request and streaming behavior.

Credentials and model sync

  • Use a project-scoped or limited provider key when the provider offers one
  • Save the provider before syncing models
  • Copy the model identifier returned by the provider rather than guessing from a display name
  • Choose a default model after sync; switch per chat when a task needs another model
  • Disconnect turns a provider off and deletes its locally saved key; if the key may have been exposed, also revoke or rotate it at the provider

Common connection failures

  • 401 or 403 — invalid, expired, incomplete, or unauthorized credential
  • 404 — incorrect Base URL, models path, wire, or model identifier
  • 429 — provider request/token quota reached
  • No usable models returned from the provider endpoint — the models endpoint returned no compatible models
  • Connection error — proxy, certificate, DNS, or local service availability problem

Implementation sources

Related pages

Ready when you are.

Kition is a desktop AI workspace for connected documents, structured tables, editable designs, whiteboards, visual workflows, and reviewable AI agents.