Connect AI Providers
Set up cloud or local AI providers and manage the models you use.
Connect AI providers
Connect a provider in the desktop app to run prompts, use AI prompt assist, or judge results.
Connect a provider
- Open Settings → AI Providers and select Connect a provider….
- Choose OpenAI, Anthropic, Google, another listed provider, or Custom OpenAI-compatible provider.
- Enter an API key, or choose Use environment key when PromptBranch finds a supported key in your environment.
- For a custom provider, enter its base URL and a model to test.
- Select Connect.
PromptBranch tests the connection before you use it. If the test fails, the provider remains available so you can correct the details and try again.
Use local models
For Ollama or LM Studio, choose Custom OpenAI-compatible provider and enter the local server's OpenAI-compatible base URL. Then open Manage models in AI Providers and add the model IDs you have installed or loaded. They will appear in the run picker.
Manage models
Use Manage models to hide models you do not use, restore hidden models, or add model IDs for local endpoints. Standard providers show models from the locally saved catalog when available.
Your API keys
PromptBranch stores provider keys encrypted with your operating system's secure storage. Keys are not included in library exports or copied to paired devices, so configure them separately on each device.
Model runs and catalog refreshes need a network connection. Browsing and editing your prompt library do not.