Providers

Set up OpenAI

Connect an OpenAI API key, sync the models available to it, and use the Responses wire in Kition.

Updated

By Kition Docs

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

Before you connect

Create or choose an OpenAI project and issue an API key for it. Provider access, billing, model availability, and rate limits are controlled by OpenAI rather than by your Kition plan.

Use a dedicated key with the smallest practical scope and provider-side spend controls.

Connect OpenAI

  • Open Settings → AI Providers and select OpenAI
  • Paste the key into the API key field
  • Leave Models endpoint under Advanced as /models unless your account or gateway documents a different path
  • Choose Sync now to fetch models, enable OpenAI, and make it the active provider
  • Open the Agent model picker and choose a text-generation model returned for your key

Note

Save provider stores edits, but a successful Sync now is what enables the provider, records its model list, and makes it active.

Connection used by Kition

Kition uses the OpenAI Responses API for the OpenAI provider and asks OpenAI not to store responses. It does not fall back to Chat Completions when a Responses request fails.

text
Base URL:        https://api.openai.com/v1
Models endpoint: /models
Wire API:        responses
Authentication:  Bearer API key

Choose a model from the synced list

Do not copy a model ID from an old guide. Sync the list for your key and select a current text-generation model that your account can use.

If no model was chosen yet, Kition selects the first returned entry. If that entry is not suitable for Agent text work, change it in the Agent model picker before sending a task.

Credential storage

Kition stores the key outside the workspace, as a local plaintext file in the secrets folder of its application data. The operating-system keychain is not used.

Heads up

Protect the operating-system account and disk, exclude Kition application data from shared backups, and revoke the key in OpenAI immediately if the device or key may be compromised.

Troubleshoot safely

  • 401 or 403 — verify the key and its project access
  • 404 during Sync now — restore the built-in /models path
  • 429 — review the project’s quota, rate limits, and billing state in OpenAI
  • A model appears but generation fails — confirm that the selected model supports the Responses API and tool calls
  • Unexpected usage — revoke the key at OpenAI and issue a replacement; Disconnect only deletes the copy saved on this device

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.