Anchor Node
One endpoint. Every model.
The gateway speaks both the OpenAI and the Anthropic protocol. You hold one key; it holds the provider keys, on the server.
01
Point your tool at it
Two addresses. Pick the one your client speaks.OpenAI protocol
opencode · aider · Codex · Cursor · Open Design · anything taking an OpenAI base URLhttps://api.hjc925.cloud/v1
Anthropic protocol
Claude Code and other Anthropic-protocol clientshttps://api.hjc925.cloud
Auth variable
Anthropic clients authenticate with ANTHROPIC_AUTH_TOKEN — not
ANTHROPIC_API_KEY. The wrong variable fails as if the key were invalid.
02
Documentation
Open to anyone with the link.
Configure your agent
Copy-paste settings, model names, and a curl test for both protocols.
Get your key
Keys are issued per person with a cost limit. Ask Herman for yours.
Models, pools and weights
How one public name is served by several providers, what a weight actually does, and how a provider that runs out of quota is held out of rotation.
What the gateway is
The two addresses, why the gateway exists, and how it keeps provider keys on the server.
API reference
Every endpoint the gateway exposes, rendered live from its OpenAPI spec.
03
Administration
Sign-in required. Not documentation.
Sign in
Dashboard
Keys, teams, models, budgets and spend. LiteLLM's own admin UI.
Sign inCompanion
Edits config.yaml, the router and the provider list. Where pool weights are set.
Sign inPool status
Which members of each pool are serving, the weight each carries, and where traffic goes.
The api.hjc925.cloud host refuses admin requests — it serves model traffic only.
Everything on this band lives on llm.hjc925.cloud.