intellibrix v1.0.0
    Preparing search index...

    Type Alias ToolDefinition<In, Out>

    A typed tool a Provider can be offered and the agent loop can execute.

    type ToolDefinition<In = any, Out = any> = {
        description: string;
        execute: (input: In, ctx: ToolContext) => Promise<Out>;
        input: StandardSchemaV1<In>;
        name: string;
        needsApproval?:
            | boolean
            | ((input: In, ctx: ToolContext) => boolean | Promise<boolean>);
        output?: StandardSchemaV1<Out>;
        resultTrust?: Trust;
    }

    Type Parameters

    • In = any
    • Out = any
    Index
    description execute input name needsApproval? output? resultTrust?
    description: string
    execute: (input: In, ctx: ToolContext) => Promise<Out>
    input: StandardSchemaV1<In>

    A Standard Schema (e.g. a Zod 4 schema) describing the tool's input.

    name: string
    needsApproval?:
        | boolean
        | ((input: In, ctx: ToolContext) => boolean | Promise<boolean>)

    Whether this tool requires approval before executing. May be a static boolean or a predicate evaluated per-call (e.g. keying off ctx.trust).

    output?: StandardSchemaV1<Out>

    An optional Standard Schema describing the tool's output.

    resultTrust?: Trust

    Trust level stamped onto this tool's result Parts. Tools returning content fetched from outside the process (MCP, workspace reads, GitHub) should set 'untrusted'.

    Settings
    Member Visibility

    On This Page
    Properties