Tool configuration
GitHub Copilot's custom model support varies by version, IDE, and organization policy. If your Copilot environment exposes a custom model entry, you can connect it to SeaLink with an API Key and the OpenAI-compatible endpoint.
Custom model support varies across Copilot versions, IDEs, and GitHub organization policies. To use SeaLink, your Copilot environment needs to expose a custom model or custom endpoint setting. If your current setup doesn't support this yet, don't worry — we have full integration guides for Cursor, Continue, Cline, and other tools that offer more complete custom model support.
VS Code: press Ctrl+Shift+P (Cmd+Shift+P on Mac), type Preferences: Open User Settings (JSON), and select it. JetBrains: look for custom model or custom endpoint options in the Copilot plugin's settings panel.
If your Copilot environment supports settings.json configuration, add the following: "github.copilot.advanced": { "customModel": { "baseUrl": "https://api.sealink.io/v1", "apiKey": "$SEALINK_API_KEY", "model": "claude-opus-5-5" } } This uses your SeaLink API Key. We recommend starting with claude-opus-5-5 to verify connectivity, then switching to your preferred model once everything checks out.
After saving settings.json, restart your IDE. Once it's back up, send a short message in Copilot Chat to test. If your Copilot environment routes requests through the custom model, you should see a response from the SeaLink model. A good first test is something like 'introduce yourself in one sentence' — quick and easy to verify.
If you can't find a custom model setting in your Copilot environment, or the config doesn't seem to take effect — this isn't a problem with your account or setup. GitHub's custom model feature availability varies by plan and version. We recommend trying one of these alternatives, which all have first-class custom model support: Cursor, Continue, Cline, Codex CLI, or Claude Code. Full setup guides for each are available on this site.
No custom model setting visible
This likely means your Copilot version or organization policy doesn't expose custom models yet. This isn't an issue with your setup — custom model availability varies across Copilot plans and versions. We suggest trying Cursor, Continue, or Cline as alternatives. They have full custom model support and the setup process is more straightforward.
Copilot still uses the default model after configuration
The Copilot client or your organization policy isn't picking up the custom model config. Double-check that the setting entry exists and the format is correct, then restart your IDE. If the issue persists, switching to Cursor, Continue, or Cline is a reliable fallback — all have been thoroughly tested with SeaLink.
Authentication failed (401 or 403)
Check two key fields: make sure apiKey is your full SeaLink API Key (starting with sk-sealink-), and baseUrl ends with /v1. Don't append /chat/completions or any other path — Copilot handles that for you.
Model call fails or returns unexpected results
Start with claude-opus-5-5 as a baseline. Avoid using Claude models with the OpenAI-compatible config, since they use a different protocol. If claude-opus-5-5 works but other models don't, check the /models page to confirm each model's status and compatibility.
