UNPKG

@postman/postman-mcp-server

Version:

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)

62 lines (61 loc) • 2.63 kB
import { z } from 'zod'; import { asMcpError, McpError } from './utils/toolHelpers.js'; export const method = 'getApiCatalogSystemEnvironmentAssociations'; export const title = "Get an API Catalog system environment's associations"; export const description = 'Lists the workspace environments attached to a system environment. Use this to see\nwhich Postman environments feed a deployment stage before adding or removing any.\nNarrow to one workspace with \\`workspaceId\\`; page with \\`limit\\` (max 100) and \\`cursor\\`.\nRequires a Postman Enterprise plan.\n'; export const parameters = z.object({ systemEnvironmentId: z.string().describe("The system environment's ID."), workspaceId: z .string() .describe("The workspace's ID to which the environments in the association belong to.") .optional(), limit: z .number() .int() .gte(1) .lte(100) .describe('The maximum number of rows to return in the response, up to a maximum value of 100. Any value greater than 100 returns a 400 Bad Request response.') .default(20), cursor: z .string() .describe('The pointer to the first record of the set of paginated results. To view the next response, use the `nextCursor` value for this parameter.') .optional(), }); export const annotations = { title: "Get an API Catalog system environment's associations", readOnlyHint: true, openWorldHint: false, destructiveHint: false, idempotentHint: true, }; export async function handler(args, extra) { try { const endpoint = `/api-catalog/system-environments/${encodeURIComponent(String(args.systemEnvironmentId))}/associations`; const query = new URLSearchParams(); if (args.workspaceId !== undefined) query.set('workspaceId', String(args.workspaceId)); if (args.limit !== undefined) query.set('limit', String(args.limit)); if (args.cursor !== undefined) query.set('cursor', String(args.cursor)); const url = query.toString() ? `${endpoint}?${query.toString()}` : endpoint; const options = { headers: extra.headers, }; const result = await extra.client.get(url, options); return { content: [ { type: 'text', text: `${typeof result === 'string' ? result : JSON.stringify(result, null, 2)}`, }, ], }; } catch (e) { if (e instanceof McpError) { throw e; } throw asMcpError(e); } }