TypeScript interfaces for MCP tools implementation. Defines parameter types, result structures, and error handling for tools.

Core Purpose

• Define tool parameter types
• Structure tool results
• Standardize error handling
• Enable type-safe tools

Tool Parameters

Scan Content

interface ToolsScanContentArgs {
    prompt?: string
    response?: string
    context?: string
    profileName?: string
    profileId?: string
    metadata?: {
        appName?: string
        appUser?: string
        aiModel?: string
        userIp?: string
    }
}

Scan Async

interface ToolsScanAsyncArgs {
    requests: Array<{
        reqId: number
        prompt?: string
        response?: string
        context?: string
        profileName?: string
        profileId?: string
    }>
}

Get Results

interface ToolsGetScanResultsArgs {
    scanIds: string[]
}

interface ToolsGetThreatReportsArgs {
    reportIds: string[]
}

Integration in Application

• Used By: Tool handler implementations
• Pattern: Args validation and typing
• Results: Consistent response format
• Errors: Structured error handling

Result Types

Tool Result Content

interface McpToolResultContent {
    type: 'text' | 'resource'
    text?: string
    resource?: McpResourceContent
}

Success Pattern

{
    content: [{
        type: 'text',
        text: 'Scan completed. Category: safe'
    }, {
        type: 'resource',
        resource: {
            uri: 'airs://scan-results/123',
            mimeType: 'application/json',
            text: JSON.stringify(data)
        }
    }]
}

Error Pattern

{
    content: [{
        type: 'text',
        text: 'Error: Invalid input'
    }],
    isError: true
}

Type Safety Benefits

Parameter Validation

// At least one content field required
if (!args.prompt && !args.response && !args.context) {
    throw new Error('Content required');
}

Profile Resolution

// Type-safe profile selection
const profile = args.profileName 
    ?? args.profileId 
    ?? config.defaultProfile
    ?? 'Prisma AIRS';

Result Construction

// Type-safe result building
const result: McpToolsCallResult = {
    content: contents,
    isError: false
};

Usage Example

// Implement tool with types
async function scanContent(
    args: ToolsScanContentArgs
): Promise<McpToolsCallResult> {
    // Validate typed arguments
    const request: AirsScanRequest = {
        ai_profile: { 
            profile_name: args.profileName 
        },
        contents: [{
            prompt: args.prompt,
            response: args.response
        }]
    };
    
    // Return typed result
    return {
        content: [{
            type: 'text',
            text: 'Scan complete'
        }]
    };
}

Related Modules

• Tools Handler - Tool implementations
• MCP Types - Protocol types
• AIRS Types - API types