Bring your own provider
BIGGI connects from its local runtime to the model provider you choose. The provider owns authentication, request processing, quotas, and billing.
Connection options
| Method | Best for |
|---|---|
| API key | Most hosted providers, including Anthropic, OpenAI, Gemini, Mistral, and OpenRouter |
| Provider login | Supported subscription or OAuth connections, such as ChatGPT and GitHub Copilot |
| Environment variables | Existing local development setups and managed machines |
| Custom provider | OpenAI-compatible or other supported endpoints with a custom base URL and model list |
| Local provider | Ollama, LM Studio, and compatible local servers |
Connect from the extension
- Open BIGGI → Settings → Providers.
- Select a provider or choose Custom provider.
- Complete the requested credential fields or login.
- Open Models and select a model.
Connections apply to all BIGGI panels in the current VS Code extension host.
Configuration files
BIGGI also reads global and project configuration:
- Global:
~/.config/biggi/biggi.jsonc - Project:
biggi.jsonc - Project alternative:
.biggi/biggi.jsonc
Use the extension settings for ordinary setup. Edit JSONC when you need provider options, environment-driven configuration, or a configuration you can review as text.
⚠️Warning
Do not commit API keys. Use the provider screen, a local environment variable, or an ignored local configuration file.
Data boundaries
- Source files and session history are local by default.
- Prompts, attachments, and tool results sent to a hosted model are processed by that provider.
- Local models keep inference local, subject to the local server you configured.
- Remote MCP servers, web tools, browser automation, and shared-session links have separate data boundaries.
Billing
BIGGI does not currently sell managed credits or subscriptions. Review pricing, budgets, and usage in the selected provider's dashboard. See Usage and Costs.