Continue.dev

Continue.dev is an open-source AI coding assistant for VS Code and the JetBrains family. It's configured via a YAML file, which makes it easy to wire up several TokOne models at once.

Install

VS Code

Search "Continue" in the marketplace and install.
Homepage: https://marketplace.visualstudio.com/items?itemName=Continue.continue

JetBrains

Settings → Plugins → Marketplace, search "Continue" and install.
Homepage: https://plugins.jetbrains.com/plugin/22707-continue

Configure

Edit ~/.continue/config.yaml (%USERPROFILE%\.continue\config.yaml on Windows):

models:
  - name: TokOne GPT-5.5
    provider: openai
    model: gpt-5.5
    apiBase: https://api.tokone.ai/v1
    apiKey: sk-xxx
  - name: TokOne GPT-5.4
    provider: openai
    model: gpt-5.4
    apiBase: https://api.tokone.ai/v1
    apiKey: sk-xxx

TIP

YAML indentation is sensitive — align all fields to the same column as name:. provider must be lowercase openai.

Field reference

Field	Meaning
name	Display name in the Continue model dropdown; pick your own
provider	Must be openai (lowercase)
model	TokOne Model ID, all lowercase, e.g. gpt-5.5
apiBase	https://api.tokone.ai/v1
apiKey	your sk-xxx

Continue reloads the config automatically after saving — no editor restart needed.

Verify

Open the Continue sidebar, select "TokOne GPT-5.5" from the model dropdown, and send a test message. The request appears on the console "Usage" page.

FAQ

"unsupported provider"

The provider field must be lowercase openai, not OpenAI / OPENAI.

YAML changes don't take effect

• Check the YAML indentation — all fields should align to the same column as name:.
• Restart VS Code / JetBrains to force a config reload.
• Check the Continue Output panel for YAML parse errors.

apiBase vs baseUrl

Continue.dev uses apiBase, not the OpenAI SDK's base_url or Cursor's "Base URL". The wrong field name is silently ignored.

Configuring a Tab autocomplete model too

Continue.dev supports a separate autocomplete model via the tabAutocompleteModel field in ~/.continue/config.yaml. Autocomplete is latency-sensitive, so the cheaper, faster gpt-5.4 is a better fit.

Next steps

• Models — model pricing
• API Endpoints — protocol notes