Skip to content

halo onboard local

My machine: create (or keep) halos.yaml, validate it and plan the managed-config install

Defaults come from this machine: –tools pins every CLI on PATH at its installed version, –provider is the one whose key is exported, and the gateway is a local halo-proxy on http://127.0.0.1:8088. An existing halos.yaml is never overwritten. –apply writes halos.yaml only; managed config is written by halo onboard install --apply.

Terminal window
halo onboard local [flags]
Flag Shorthand Type Default Description
--apply bool false write halos.yaml (default: dry run)
--gateway string gateway base URL the CLIs call (default: local halo-proxy http://127.0.0.1:8088)
--gateway-engine string what serves –gateway: halo-proxy (default) | kong | external (your company’s own API gateway: CLIs get provider model ids)
--model stringArray [] model alias=provider model id, repeatable
--org string local organization name
--policy-dir string policy repo directory (default “.”)
--project string Google Cloud project (provider vertex)
--provider string anthropic | bedrock | vertex | openai | gemini | multi (default: from the exported key)
--ring string ring this machine follows (default: the GA ring)
--rollout string standard fast | standard | careful
--root string plan the install under this directory instead of the real paths
--safety string standard strict | standard | relaxed
--tools stringSlice [] harnesses as name@version (default: every CLI on PATH at its installed version)
Flag Shorthand Type Default Description
--output string text output format: text|json

halo onboard