Run supported models locally and route other requests through a multi-provider gateway with Conifer.
Conifer connects local model execution with a cloud inference gateway. Developers can run supported weights on their own hardware or send requests through a shared interface to hosted models. Routing can choose where a request runs, while an explicit model choice gives the caller more direct control over the result and provider path.
The gateway works with existing OpenAI and Anthropic clients when configured with the appropriate endpoint and credentials. Conifer also provides a command-line runtime and a macOS desktop application. The official platform details distinguish full local engine support on Apple Silicon and Windows from the Linux thin client, which connects to the gateway or another Conifer host. Windows and Linux desktop applications are not presented as currently available.
The model catalogue includes frontier providers and open-weight families. Local execution depends on the engine’s support and available hardware, so a catalogue entry should not be taken as proof that every model can run on a particular computer. The CLI manages model installation and local serving, giving other tools an endpoint for accessing the runtime.
Conifer describes query-based routing that considers the needs of each request. Users can also bring their own provider keys or connect compatible endpoints. A local request and a cloud request have different privacy properties: local inference keeps its prompt and response on the machine, while cloud execution sends the required context to the chosen endpoint. Pinning a route to local is therefore different from merely allowing local execution as one option.
The privacy documentation separates usage telemetry from optional content sharing. Teams should review those controls before processing sensitive code or documents. A routing policy should be tested with the actual client, model and credentials used in production.
The runtime, router and local inference are advertised as free. Cloud usage is prepaid and charged at the model’s rate without a Conifer markup; bringing provider keys does not remove the provider’s own charges. Start with a small known task, inspect the selected route and compare quality, latency and cost before enabling automatic routing for broader workloads.
No other apps in this category yet.
