ToolMessage

@open-agent-loops/core

Defined in: types/tool-message.ts:37

A tool-result turn — the response to a single ToolCall.

Remarks

The loop produces exactly one ToolMessage per tool call the assistant made, carrying that tool's output in content. It is linked back to the call it answers by ToolMessage.tool_call_id (which equals the call's id), so the model can tell which result belongs to which call. N tool calls in a turn yield N tool messages, paired one-to-one by id.

Abstraction over — OpenAI's tool message ({ role, tool_call_id, content }); toolName and isError are this loop's extensions (plus timestamp from the base). Egress drops the extensions, so only the three standard fields cross the wire (see toChatMessage).

Example

Wire shape (OpenAI chat-completions)

{
  "role": "tool",
  "tool_call_id": "call_abc123",
  "content": "72°F and sunny"
}

See

ToolCall

Extends

• MessageBase

Properties