Model providers in Aldwin
Pick where Aldwin's model runs, from a free model on your own computer to a cloud API key.
Aldwin needs one model to think with. The model can run on your own computer, or with a cloud company you have an API key for. You can switch between them at any time, in the middle of a conversation.
Providers#
| Provider | Runs | Cost | Guide |
|---|---|---|---|
| Ollama | On your computer | Free | Set up Ollama |
| LM Studio | On your computer | Free | Set up LM Studio |
| OpenAI | Cloud | Paid, prepaid credits | Set up OpenAI |
| Anthropic | Cloud | Paid | Set up Anthropic |
| Google Gemini | Cloud | Free tier with limits, then paid | Set up Gemini |
| OpenRouter | Cloud | Free models with daily limits, paid models with credits | Set up OpenRouter |
| DeepSeek | Cloud | Paid, prepaid balance | Set up DeepSeek |
| Other server | Wherever it runs | Depends on the server | Connect another server |
Not sure where to start? A local model costs nothing and keeps your prompts on your computer, but it needs a fairly recent machine. A cloud model works on any computer and is usually smarter, but you pay per use.
Connect a provider#
In the web app, open Settings and choose Models. Running local providers show under On this computer. For a cloud provider, select Add key and paste your key. Aldwin checks it with the provider before saving it. Model in use picks the model, and you can also switch from the message box in any chat. See Web app settings.
In the terminal, type /model to open the provider list.
- Local providers (Ollama, LM Studio) appear on their own when they are running. Pick one, then pick a model.
- Cloud providers need a key first. Pick Add a provider…, choose the company, and paste your key. Aldwin checks the key with the provider before saving it.
- Manage keys… updates or removes a saved key.
On first launch, aldwin and aldwin web both walk you through the same choice. Run aldwin setup to go through it again.
If you know the model name, you can skip the list: type /model provider/model-name, for example /model ollama/qwen3:8b.
Choosing a model#
The terminal's model list shows two things next to each name:
✓ toolsmeans the model can use Aldwin's tools, such as reading files, editing, searching and running commands.✗ chat onlymeans it can only talk. Aldwin warns you before switching to a chat-only model.- A number like
128Kis the context window: how much conversation the model can hold at once.
Small local models sometimes answer in text and skip tool steps. If edits or searches never happen, try a bigger tool-capable model.
The status line at the bottom of the terminal shows how much context is left. In the web app, the ring next to the model picker shows it. When it runs low, older parts of the conversation drop out. Type /new to start fresh; memory notes you saved stay.
Where keys are kept#
Keys you paste in Aldwin are saved in ~/.aldwin/secrets.yaml, which only your user account can read. Each guide also lists an environment variable (for example OPENAI_API_KEY). An environment variable wins over a saved key.
Where your prompts go#
- Local providers: to the model on your computer. Nothing leaves your machine.
- Cloud providers: to that company's servers over the internet, under its own terms. Each guide has a short privacy note.
Files Aldwin reads for you are part of the prompt, so they go wherever the model is. More in Privacy.