AI Providers Overview
What are AI Providers?
By default, every conversation in Adri runs on Adri AI — the Adri-managed model. You do not have to configure anything.
If you want conversations to run on your own LLM account instead, add an AI Provider. A provider holds the connection to an LLM backend: its type, base URL, encrypted API key, and the models it offers.
Once a provider is added:
- Providers are shared across the project — every project member can pick them.
- Each conversation picks a provider and one of its models from the model picker in the chat composer.
- The picker is available in the Adri web app and in the Adri AI Foundry VS Code extension.
Supported provider types
| Provider type | Status | Where to get an API key |
|---|---|---|
| Anthropic (Claude) | Available | Anthropic Console → API keys |
| OpenAI (GPT) | Available | OpenAI Platform → API keys |
| Google (Gemini) | Coming soon | — |
| Azure OpenAI | Coming soon | — |
| OpenAI-compatible | Coming soon | — |
| AWS Bedrock | Coming soon | — |
| LiteLLM proxy | Coming soon | — |
Today you can connect Anthropic (Claude) and OpenAI (GPT). The remaining types are on the roadmap — this page will be updated as they become available.
How the default "Adri AI" entry works
The model picker always shows Adri AI as the first entry:
- Adri AI = "use the project default". The conversation inherits the project's default provider and model. If you never added a provider, this is the Adri-managed model.
- Picking a specific model pins this conversation to that provider and model.
- Switching takes effect on the conversation's next message — earlier messages are never rewritten.
Where do the model lists come from?
You never type model names by hand. For each provider, Adri lists the available models automatically from a built-in catalog plus live discovery against your endpoint. Each model appears with a friendly label (for example Claude Sonnet 4.5).
Security
- API keys are stored encrypted and are write-only — after saving, a key is never shown again, not even to you. Adri only reports whether a key is configured.
- Conversations that use a provider call the model with your key — usage is billed to your provider account.
Set it up
- Add a provider — connect your LLM account (5 minutes).
- Manage providers — set a default, rotate keys, remove providers.
- Switch models in web chat — pick a model per conversation.
- Switch models in VS Code — the same picker inside the extension.
- FAQs & troubleshooting — quick answers when something looks off.