Continue.dev OpenAI-Compatible API Setup

Continue.dev uses config.yaml to manage models. Its official config reference supports model fields like provider, model, apiBase, and roles, which makes it suitable for OpenAI-compatible APIs.

Where to configure

Newer Continue setups use config.yaml. You can edit it from Continue's configuration UI or migrate from the older config.json format following the official docs.

Basic config example

name: Cheng Zi AI
version: 1.0.0
schema: v1
models:
  - name: GPT Coding
    provider: openai
    apiKey: sk-your-api-key
    apiBase: https://api.cheng-zi-ai.com/v1
    model: gpt-5.3-codex
    roles:
      - chat
      - edit
      - apply

apiBase overrides the default API base. OpenAI-compatible gateways usually use a /v1 endpoint.

Field meanings

• name: display name inside Continue.
• provider: use openai for OpenAI-compatible services first.
• apiKey: your API key.
• apiBase: custom API base URL.
• model: actual model ID.
• roles: whether the model is used for chat, edit, apply, and related workflows.

Start with one model

Do not configure many models at once. Add one model, verify chat and edit work, then duplicate the block for other models.

Common errors

Config saved but not active

Restart the IDE or Continue extension. Workspace and global configs can override each other.

401

Check that apiKey is complete and not broken by YAML indentation or quotes.

404

Check whether apiBase includes /v1. Continue's apiBase examples are usually full endpoints.

Agent tools do not work

If Agent mode needs tool use, the model may need explicit capabilities or a tool-capable model. Test normal chat first, then Agent mode.

Practical advice

Continue is best for codebase reading, editing, and agent workflows. Use cheaper models for everyday chat and stronger models for complex refactoring.