Documentation / @agentick/core / ToolMetadata

Interface: ToolMetadata<TInput, TOutput>

Defined in: core/src/tool/tool.ts:705

Type Parameters

TInput

TInput = any

TOutput

TOutput = any

Properties

aliases?

optional aliases: string[]

Defined in: core/src/tool/tool.ts:796

Alternative names for user dispatch (e.g. ["mount"] for "add-dir").

---

audience?

optional audience: ToolAudience

Defined in: core/src/tool/tool.ts:793

Controls who sees this tool. "model" (default) = model only, "user" = dispatch only, "all" = both.

---

confirmationMessage?

optional confirmationMessage: string | (input) => string

Defined in: core/src/tool/tool.ts:762

Message to show user when requesting confirmation. Can be a string or a function that receives the input. Default: "Allow {tool_name} to execute?"

---

confirmationPreview()?

optional confirmationPreview: (input, deps) => Promise<Record<string, unknown>>

Defined in: core/src/tool/tool.ts:766

Async preview metadata for confirmation UI (e.g. diff).

Parameters

input

any

deps

any

Returns

Promise<Record<string, unknown>>

---

defaultResult?

optional defaultResult: ContentBlock[]

Defined in: core/src/tool/tool.ts:744

Default result to use when requiresResponse is false. Returned immediately for render tools that don't need client feedback. Default: [{ type: 'text', text: '[{tool_name} rendered on client]' }]

---

description

description: string

Defined in: core/src/tool/tool.ts:707

---

displaySummary()?

optional displaySummary: (input) => string

Defined in: core/src/tool/tool.ts:764

Short summary string for tool call indicators.

Parameters

input

any

Returns

string

---

input

input: ZodSchema<TInput>

Defined in: core/src/tool/tool.ts:708

---

intent?

optional intent: ToolIntent

Defined in: core/src/tool/tool.ts:724

Tool intent describes what the tool does (render, action, compute). Used by clients to determine how to render/handle tool calls. Default: COMPUTE

---

libraryOptions?

optional libraryOptions: LibraryToolOptions

Defined in: core/src/tool/tool.ts:780

Library-specific tool configurations. Keyed by library name (e.g., 'ai-sdk', 'langchain', 'llamaindex'). Used by adapters for library-specific tool behavior (timeouts, callbacks, etc.). Each adapter can extend this type using module augmentation.

---

mcpConfig?

optional mcpConfig: object

Defined in: core/src/tool/tool.ts:785

MCP-specific configuration (only relevant if type === 'mcp'). Contains connection info and MCP server details.

Index Signature

[key: string]: any

serverName?

optional serverName: string

serverUrl?

optional serverUrl: string

transport?

optional transport: "websocket" | "sse" | "stdio" | "in-process"

---

name

name: string

Defined in: core/src/tool/tool.ts:706

---

output?

optional output: ZodSchema<TOutput>

Defined in: core/src/tool/tool.ts:713

Optional Zod schema for output validation. Used for type-safe tool composition and workflow orchestration.

---

providerOptions?

optional providerOptions: ProviderToolOptions

Defined in: core/src/tool/tool.ts:773

Provider-specific tool configurations. Keyed by provider name (e.g., 'openai', 'google', 'anthropic'). Preserved when converting to ToolDefinition. Each adapter can extend this type using module augmentation.

---

requiresConfirmation?

optional requiresConfirmation: boolean | (input) => boolean | Promise<boolean>

Defined in: core/src/tool/tool.ts:756

Whether execution requires user confirmation before running. Applies to any tool type (SERVER, CLIENT, MCP).

• boolean: Always require (true) or never require (false)
• function: Conditional - receives input, returns whether confirmation needed. Can be async to check persisted "always allow" state. Use Context.get() inside the function to access execution context.

Default: false

---

requiresResponse?

optional requiresResponse: boolean

Defined in: core/src/tool/tool.ts:732

Whether execution should wait for client response. Only applicable for CLIENT type tools.

• true: Server pauses and waits for tool_result from client (e.g., forms)
• false: Server continues immediately with defaultResult (e.g., charts) Default: false

---

timeout?

optional timeout: number

Defined in: core/src/tool/tool.ts:738

Timeout in milliseconds when waiting for client response. Only applicable when requiresResponse is true. Default: 30000 (30 seconds)

---

type?

optional type: ToolExecutionType

Defined in: core/src/tool/tool.ts:718

Tool execution type. Determines how the tool is executed. Default: SERVER (engine executes tool.run on server).

---

ui?

optional ui: object

Defined in: core/src/tool/tool.ts:802

MCP App UI metadata. When present, tool calls emit ui on the tool_result_start event, signaling the host to mount an app iframe.

resolveContent()?

optional resolveContent: () => Promise<string | undefined>

Optional resolver for app HTML content — see ToolMetadata.ui.resolveContent

Returns

Promise<string | undefined>

resourceUri?

optional resourceUri: string

visibility?

optional visibility: ("model" | "app")[]