Providers and models
Connect your own provider credentials, choose models per run, and understand how provider availability works.
Provider configuration is explicit. Choose the provider and model for each agent run. LegionCode does not resell model access.
Supported providers
| Provider | Status | Notes |
|---|---|---|
| OpenRouter | Supported | Use an OpenRouter API key |
| OpenAI | Supported | Use an OpenAI API key |
| Groq | Supported | Use a Groq API key |
| Anthropic | Supported | Use an Anthropic API key |
| Google Gemini | Supported | Use a Gemini API key |
| OpenCode | Supported | Use an OpenCode-compatible provider key |
| Together AI | Supported | Available where enabled for your account |
| Cerebras | Supported | Available where enabled for your account |
| Cloudflare AI | Coming soon | Not available for private-alpha runs yet |
Provider availability can vary by account, region, model access, and upstream API changes.
Model catalogs
LegionCode shows models available through the connected provider configuration. A model appearing in a provider catalog does not guarantee that your provider account can use it for every task.
If a model is missing, check the provider console first. Some providers require separate model access, billing activation, or regional availability before a model can be used.
Choosing a model
Use a model that matches the task:
- Small code edits and documentation updates usually work best with fast, lower-cost models.
- Multi-file refactors, debugging, and review-heavy tasks benefit from stronger reasoning models.
- Tool-heavy tasks require models and provider routes that support the tool protocol used by the workspace.
If execution is rejected, choose another model from the same provider or connect a provider with the required capabilities.