Connection

The agent-connection config node stores the provider endpoint and credential used by an agent. It can be shared by multiple agents.

Properties

Property	Requirement	Purpose
Name	Optional	Friendly label used in configuration selectors.
Base URL	Required at runtime	Root URL of the provider’s OpenAI-compatible API.
API Key	Required at runtime	Credential sent to the provider.

Example base URLs include a provider API root ending in /v1. Use the value documented by your provider rather than a model-specific URL.

Provider Requirements

The current agent runtime expects an OpenAI-compatible chat completions API. Depending on the selected reasoner and model, the provider must also support:

• JSON object response formatting.
• Image content when requests include images.
• Usage fields for token accounting.
• Cost fields when using spend budgets.

Compatibility can differ by model on the same provider.

The runtime currently describes skills and actions in prompts and parses structured responses. Native tool or function calling support is not required.

Shared Connections

Both Thinker stages use the connection selected on the agent. A separate Schema Model does not currently select a separate provider connection.

Model config nodes contain model settings only. They run through the Connection selected by the Agent node.

Because config nodes are reusable, changing a shared Base URL or API Key affects every agent that references the connection.

Agent Studio Editing

Agent Studio currently exposes the Name and Base URL but not the API Key. Add or change credentials in the Node-RED connection editor. If Studio and Node-RED show different values, treat the deployed Node-RED config node as authoritative.

Security

The API key is stored as a Node-RED credential. Still:

• Do not include real keys in screenshots, examples, exports, or support messages.
• Restrict access to the Node-RED editor and deployed credentials.
• Rotate a key that may have been exposed.