Integrations

Continue → AI24X

Continue is an open-source coding assistant for VS Code and JetBrains. Point it at AI24X with the OpenAI provider—one key for flash/pro and VIP named models.

1) Prerequisites

Sign up → Console → create an API key → keep a small credit balance. Install the Continue extension.

2) config.yaml (recommended)

Edit ~/.continue/config.yaml (or the config opened from Continue settings):

name: AI24X version: 1.0.0 schema: v1 models: - name: AI24X Flash provider: openai model: flash apiBase: https://api.ai24x.com/v1 apiKey: YOUR_AI24X_API_KEY roles: - chat - edit - apply - name: AI24X Pro provider: openai model: pro apiBase: https://api.ai24x.com/v1 apiKey: YOUR_AI24X_API_KEY roles: - chat - edit

3) config.json (legacy)

{ "models": [ { "title": "AI24X Flash", "provider": "openai", "model": "flash", "apiBase": "https://api.ai24x.com/v1", "apiKey": "YOUR_AI24X_API_KEY" } ] }

4) Pitfalls

  • apiBase must include /v1.
  • 401 = bad, truncated, or rotated key—paste the full key, or create a new one in Console.
  • Start with flash; VIP named ids need an active membership.
  • If a build defaults to Responses: text chat can use /v1/responses; for tools / agent tool-calling set useResponsesApi: false and use chat/completions.

Verify the key outside the editor

Before touching Continue settings, confirm the key and base URL with the curl guide from a terminal. If curl returns 200, any remaining problem lives in the config file; if it returns 401, the key is the problem no matter what the editor shows.

Splitting roles across tiers

The roles list on each model entry decides which Continue features use it. A practical split: give flash the high-frequency roles and keep pro on chat and edit, where reasoning matters most. Autocomplete fires far more often than chat, so leaving it on the cheaper tier is the single biggest saver.

Because both entries share one key and one base URL, moving a role between tiers is a two-line config change — no new account, no new key.

Troubleshooting order

  1. Verify the endpoint with curl; expect HTTP 200.
  2. Confirm apiBase ends at /v1.
  3. Confirm the key in the config file is complete and current.
  4. Reload the Continue window so the new config is picked up.
  5. If tool-calling misbehaves, set useResponsesApi: false as noted above and retry.

Create account LiteLLM guide Cursor guide Console