@postman/postman-mcp-server/dist/src/tools/submitContextGraphAsk.js

Postman MCP Server — connect AI agents (Claude Code, Cursor, VS Code Copilot, Gemini CLI) to your Postman collections, specifications, and environments via Model Context Protocol (MCP)

import { z } from 'zod';
import { ContentType } from '../clients/postman.js';
import { asMcpError, McpError } from './utils/toolHelpers.js';
export const method = 'submitContextGraphAsk';
export const title = 'Submit a Context Graph ask';
export const description = "Asks a natural-language question about the team's software estate — which APIs and\nservices exist, what they expose, and how they depend on each other — and gets an\nanswer grounded in the team's Context Graph. Use this for questions about the estate\nas a whole that no single collection or workspace can answer.\nThis is asynchronous and does not return the answer: it returns an \\`askId\\` with\nstatus \\`pending\\`. Poll getContextGraphAsk with that ID until \\`status\\` is \\`completed\\`\nor \\`failed\\`, and only report an answer once it is \\`completed\\`. Set\n\\`includeAnswer=false\\` when you want just the graph data and intend to phrase the\nanswer yourself.\nDo not use this tool to search Postman elements by name; use searchPostmanElements\ninstead. Each ask counts against the team's quota, so ask one well-formed question\nrather than retrying variations of the same one. Requires Context Graph to be\nenabled for the team.\n";
export const parameters = z.object({
    query: z
        .string()
        .min(1)
        .max(2000)
        .describe('The natural-language question to answer against the graph.'),
    includeAnswer: z
        .boolean()
        .describe("Whether to include the prose answer in the result. If `false`, the result's `answer` property is `null` and the result's other properties are unaffected.")
        .default(false),
});
export const annotations = {
    title: 'Submit a Context Graph ask',
    readOnlyHint: false,
    openWorldHint: true,
    destructiveHint: false,
    idempotentHint: false,
};
export async function handler(args, extra) {
    try {
        const endpoint = `/context-graph/asks`;
        const query = new URLSearchParams();
        const url = query.toString() ? `${endpoint}?${query.toString()}` : endpoint;
        const bodyPayload = {};
        if (args.query !== undefined)
            bodyPayload.query = args.query;
        if (args.includeAnswer !== undefined)
            bodyPayload.includeAnswer = args.includeAnswer;
        const options = {
            body: JSON.stringify(bodyPayload),
            contentType: ContentType.Json,
            headers: extra.headers,
        };
        const result = await extra.client.post(url, options);
        return {
            content: [
                {
                    type: 'text',
                    text: `${typeof result === 'string' ? result : JSON.stringify(result, null, 2)}`,
                },
            ],
        };
    }
    catch (e) {
        if (e instanceof McpError) {
            throw e;
        }
        throw asMcpError(e);
    }
}