openai provider and the OpenAI-compatible /v3 base URL, so you only add models to your config.yaml.
Overview
Continue is an open-source code assistant for VS Code and JetBrains IDEs. Each model you add to itsconfig.yaml appears in the model dropdown, and a roles list decides whether Continue uses it for chat, edits, autocomplete or embeddings. One Eden AI key and one apiBase cover all of them.
Prerequisites
- Continue installed from your IDE marketplace: in VS Code, search for “Continue” in Extensions; in JetBrains, open Settings (Ctrl+Alt+S / Cmd+,), select Plugins and search for “Continue”
- An Eden AI API key from app.edenai.run
Setup
Step 1: Open the config file
Open the configs dropdown in the Continue panel and click the gear icon next to your local config (Main Config on a fresh install).config.yaml opens in your editor (on a fresh install it is ~/.continue/config.yaml, or %USERPROFILE%\.continue\config.yaml on Windows).
Continue 2 prefers config.yaml. If one exists, a config.json next to it is ignored, so do not edit the JSON file.
Step 2: Add Eden AI models
Replace the emptymodels: [] in the file with these entries (the file already has name, version and schema; keep them):
provider: openai makes Continue send OpenAI-style requests to apiBase. useLegacyCompletionsEndpoint: false keeps autocomplete and edits on /v3/chat/completions, because Continue otherwise calls the legacy /completions route, which Eden AI does not serve. Continue guesses whether a model supports tools from its id, and that guess fails for provider/model ids, so capabilities: [tool_use] turns on native tool calling for Agent mode.
Step 3: Store your key
The${{ secrets.EDEN_AI_API_KEY }} reference keeps the key out of the config. Continue looks it up in ~/.continue/.env, then in the .env files of your workspace, then in the process environment. Create ~/.continue/.env:
Step 4: Save
Saveconfig.yaml. Continue reloads it automatically and the new models appear in the model dropdown.
Model naming
Use the fullprovider/model id as model, such as anthropic/claude-sonnet-latest, openai/gpt-latest or openai/gpt-mini-latest. Aliases ending in -latest follow the newest model of a family. A model you use in Agent mode needs function calling; see Listing models for each model’s capabilities.
Embeddings
Theembed role in the config above sends Continue’s embedding requests to Eden AI’s /v3/embeddings. Use an id from the embeddings catalog, such as openai/text-embedding-3-small.
Expert Models as Tools (MCP)
Eden AI also hosts an MCP server that exposes the expert model catalog (OCR, web search, speech, translation and more) as tools. Add the server to your assistant’sconfig.yaml:
Troubleshooting
Models Not Appearing
If models don’t appear in the dropdown:- Check that
config.yamlis valid YAML and that each model hasname,providerandmodel. Config errors appear in the Configs section of Continue’s settings. - Ensure
apiBaseis exactly:https://api.edenai.run/v3
Authentication Errors
If you see 401 errors:- Verify the key in
~/.continue/.env(or in the.envfile yoursecretsreference reads) - Check there are no extra spaces in the API key
- A 401 with
Invalid API keymeans thesk-eden-...key is wrong or revoked;Invalid tokenmeans the credential is not an Eden AI key at all. - Test the API key manually:
Agent mode does not call tools
Addcapabilities: [tool_use] to the model, and use a model that supports function calling.
Next Steps
- Listing models - Browse available providers and models
- MCP server - Expert models as tools
- Continue config.yaml reference - Every option Continue accepts