AI Provider Settings
Configure AI providers, API keys, and model parameters.
AI Provider Settings
Configure your AI providers, API keys, models, and advanced parameters.
Provider Selection
Select from available AI providers:
| Provider | Description |
|---|---|
| OpenAI | GPT-5.6 (Sol/Terra/Luna), GPT-5.5, GPT-5.4 |
| Anthropic | Claude Opus 5, Sonnet 5, Haiku 4.5, Fable 5 |
| Google Gemini | Gemini 3.6 Flash, 3.5 Flash, 3.1 Pro |
| xAI | Grok 4.6, Grok 4.5, Grok 4.3, Grok 4.20 |
| DeepSeek | DeepSeek V4 Flash and Pro |
| Ollama | Local models via Ollama |
| Custom | Your own OpenAI-compatible endpoints |
Open the provider dropdown to select a provider โ search for it, or use โโ and Enter. The active provider is shown at the top of the list. Use the + Add button for a custom provider.
API Key
Enter your API key for the selected provider:
- Keys are stored locally in your browser
- Never sent to BraceKit servers
- Unique per provider
Paste your key into the API Key field. Click the eye icon to show/hide the value.
Note: Ollama running on localhost doesnโt require an API key.
Base URL
For Custom Providers and Ollama, you can configure a custom Base URL:
- Ollama: Change from default
http://localhost:11434to connect to a remote server - Custom: Enter the full API endpoint URL
Model Selection
Choose a model for the selected provider from the searchable list. BraceKit fills the list when it can (after you enter an API key). You can also type a model name if it is not listed yet.
Managing Models (Custom Providers)
Custom providers keep a list of models you add:
- Click a row to make it the active model
- Click Add model to open a dialog (model id, optional display name, and specs)
- Click the pencil to edit a modelโs specs
- Click ร to remove a model
Custom Providers
Add your own API endpoints that are compatible with OpenAI, Anthropic, Gemini, or Ollama formats.
Adding a Custom Provider
- Click the + Add button next to the provider dropdown
- Fill in the details:
- Name: Display name for the provider
- Format: API format (OpenAI, Anthropic, Gemini, or Ollama)
- Base URL: API endpoint (e.g.,
https://api.example.com/v1)
- Click Save Provider
Removing a Custom Provider
Hover over a custom provider button in the grid and click the ร that appears in the top-right corner.
Model specs
Every model has its own limits and features โ not one setting for the whole provider. The Advanced card shows the specs for the currently selected model.
| Spec | What it affects |
|---|---|
| Context / max input / max output | How long a conversation can get (token meter and auto-compact) and how long a reply can be |
| Mode | Chat, image generation, or embeddings |
| Input / output modalities | What you can attach (text, images, PDFs) and what the model can return |
| Capabilities | Tools, vision, reasoning, structured output, Google Search, image generation |
| Reasoning control | How thinking is configured when reasoning is on |
Built-in providers: these fields are read-only. BraceKit already knows the usual limits for listed models, and updates them when the provider returns that information.
Custom providers: you set specs in the add/edit dialog and on the Advanced card.
Vision always includes Image as an input modality. You cannot turn Image off while Vision is on.
If you turn on tools or thinking in the composer (or attach a file) that this modelโs specs say it does not support, BraceKit shows a warning with Configure here โ that link opens this settings page so you can fix the spec.
Model Parameters
Fine-tune model behavior with advanced parameters. Available parameters vary by provider format.
Common Parameters
| Parameter | Description | Range |
|---|---|---|
| Temperature | Controls creativity. Low = consistent, High = varied | 0.0 - 2.0 |
| Top P | Controls word choice breadth. Lower = more focused | 0.0 - 1.0 |
| Max Tokens | Maximum response length | Any positive integer |
Advanced Parameters
| Parameter | Provider | Description |
|---|---|---|
| Top K | Anthropic, Gemini, Ollama | Limits word choices to top K options |
| Min P | Ollama | Minimum probability threshold for tokens |
| Context Window | Ollama | Context window size in tokens (num_ctx) |
| Keep Alive | Ollama | How long to keep model in memory (e.g., โ5mโ, โ24hโ) |
| Thinking Budget | Anthropic, Gemini | Max tokens for internal reasoning (when enabled) |
Resetting Parameters
Click โReset to defaultsโ to clear all custom parameter values.