Models
Models is the one module through which every agent of pcx calls a model: the agent, the chief, a team's members, and a workflow's agent and model steps. You bring the key of the provider you choose; that provider bills you, and pcx adds nothing to its price.
The page in the app: Modules and models.
Settings
Section titled “Settings”One section for each provider:
| Section | Its models |
|---|---|
| OpenRouter | every model OpenRouter routes (Claude, GPT, Gemini, Llama and more), each with its price and window |
| Anthropic | Anthropic’s Claude models |
| Z.AI | Z.AI’s GLM models, with the key of its coding plan |
| OpenAI | OpenAI’s GPT models |
Each section has:
- API key: a secret; once saved, nobody sees it again, in the app or through the API. A provider without a key is off.
- Endpoint: the URL of the provider’s API; empty, the provider’s own. Set it for a proxy or a compatible gateway.
The module runs with no key at all; it then lists no model, and its settings are where you enter one.
Which models
Section titled “Which models”- Every model a provider lists can be named. Nothing is turned on first: enter the key and its models appear.
- The Models page, and every setting that names a model (the agent’s, the chief’s, a team member’s), offer them all, each with its provider. Prices and windows show when the provider’s API gives them.
- When two providers list the same model, the first in the settings’ order answers.
- A key the provider refuses is reported on that provider’s field, and the turns that named its models end saying why.
The agent’s turns use glm-5.3 unless its settings name another model; the chief and a team’s members use the agent’s model unless theirs name one. How a turn handles a busy or failing model: Agent.