Local AI provider setup

BCCLI does not run models itself. It talks to a runtime that does, so the first step is always a runtime that is installed, running, and able to answer a request on this machine.

1. Install and start the runtime

Use the runtime you already prefer — Ollama and LM Studio are both supported, as is any other provider that speaks the OpenAI-compatible API. Install it the way that project documents, then start it.

The public CLI discovers an Ollama runtime on loopback, at http://localhost:11434. A runtime bound to a different port, or bound to a different interface, will not be found by that default.

2. Tell BCCLI it exists

bccli provider list
bccli provider add <id>

provider list shows what BCCLI already knows about. provider add registers one that is running but unknown to it.

3. Check the models arrive

bccli models

Models appear here once their provider is configured and signed in. A runtime that is running but has no models downloaded will register correctly and still contribute nothing to this list.

Local model downloads — getting a model in front of the runtime.
Use Local AI in BCCLI — pinning a local model to a run.

Local AI provider setup

1. Install and start the runtime

2. Tell BCCLI it exists

3. Check the models arrive

Next