Lab Bootstrap - Model Surface

What this adds

Choose one way to talk to a model.

This is the zero step for the tooling journey. You are not building an agent yet. You are deciding whether the rest of the work will wrap a subscription surface, direct API, managed platform endpoint, router, or local host.

The output of bootstrap is simple: one chosen path, one repeatable request shape, and a clear sentence about why you chose it.

Do this

curl "$MODEL_ENDPOINT" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input":"hello"}'

Adapt the payload to your provider or local runtime. If you already have a hosted CLI agent surface, that CLI itself can be your starting interface.

Before you put a provider key into a shell or agent host, read API key security.

Optional stronger path: put the key behind the local broker lab before you hand model access to a host.

Reference pages

Start with model access. If you want the deeper local entrance, continue with local hosting and model artifacts.

Security companion: API key security.

Done when

You can explain what endpoint or surface you are calling, what credential or runtime it depends on, and why the next lab will wrap that exact boundary.

Real-world analog

curl is the everyday version of this step: prove one request, inspect one response, and remove mystery before adding abstractions.

Skip guidance

If you already have a provider API key, managed platform deployment, or local endpoint, you can move straight to lab 00. If you are arriving from a hosted CLI assistant, lab 00 is still useful, but lab 01 may be the first genuinely new layer.

If you already store the key in a system like 1Password or Azure Key Vault, skip to secret-manager integration.

Next lab

Lab 00: start with model access.