Skip to main content
Synti talks to a model through a provider. Out of the box it defaults to Anthropic, but you can switch to another supported provider or aim a connection at any compatible endpoint you run yourself — a gateway, a proxy, or a local model server. This page covers picking a provider and setting a custom base URL. To manage several providers at once and switch between them per session, see LLM Connections.

​
Supported providers

Anything that speaks an OpenAI-compatible API — a hosted gateway, an internal proxy, or a local runtime such as Ollama — connects through the custom option with a base URL.

​
Setting a base URL

1

Open connection settings

Go to Settings → AI and add or edit a connection. On desktop you can also reach settings with Cmd/Ctrl + ,.
2

Pick a provider or choose Custom

Selecting a built-in provider fills in the right endpoint automatically. Choose Custom to enter your own base URL.
3

Enter the base URL and model

Provide the full base URL, an API key if the endpoint requires one, and the model identifier you expect to call.
4

Save

Synti validates the connection before saving so you catch a bad URL or key immediately.
You can change any of these later from the same screen without recreating the connection.

​
Model identifiers

Model naming depends on the provider:
  • Anthropic — use the model ID directly, e.g. claude-sonnet-4-5.
  • OpenRouter and other gateways — use the vendor/model form, e.g. anthropic/claude-sonnet-4-5.
  • Local runtimes — reference the model name as your server exposes it, e.g. llama3.2.
  • Custom — follow your endpoint’s documentation for valid IDs.

​
Image support on custom endpoints

Custom endpoints are treated as text-only until you tell Synti otherwise, because not every compatible server accepts images.
  • In the app — open the model picker and toggle image support for the model.
  • In config — set supportsImages at the endpoint or per model. A per-model value overrides the endpoint default.

​
Storage

Base URLs and model names are not secrets, so they live in plain text in the config file. API keys are encrypted separately in the credential store. At launch you can also override a connection’s base URL through the environment — see Environment Variables.
Local model servers reachable at an address like http://localhost:11434 usually need no API key. Make sure your runtime exposes an OpenAI-compatible route before pointing a Synti connection at it.

LLM Connections

Keep several providers and switch per session.

Credentials

Where the API key for each endpoint is stored.

Environment Variables

Override provider and base URL at launch.

Config File

Where base URLs and models are persisted.