HUMMBL MCP Server

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js"; import { z } from "zod"; import { type AuditInputItem, getSelfDialecticalMethodology, auditModelCodes, } from "../framework/self_dialectical.js"; import { isOk } from "../types/domain.js"; export function registerMethodologyTools(server: McpServer): void { // Tool: Get Self-Dialectical AI Methodology definition server.registerTool( "get_methodology", { title: "Get Self-Dialectical AI Methodology", description: "Retrieve the canonical Self-Dialectical AI Systems methodology with HUMMBL Base120 mappings.", inputSchema: z.object({}), outputSchema: z.object({ id: z.string(), title: z.string(), version: z.string(), summary: z.string(), documentUrl: z.string().optional(), totalPages: z.number().optional(), modelsReferenced: z.array(z.string()), stages: z.array( z.object({ stage: z.enum(["thesis", "antithesis", "synthesis", "convergence", "meta_reflection"]), title: z.string(), description: z.string(), modelCodes: z.array(z.string()), }) ), metaModels: z.array(z.string()), }), }, async () => { const result = getSelfDialecticalMethodology(); if (!isOk(result)) { return { content: [ { type: "text" as const, text: `Unable to retrieve Self-Dialectical methodology: ${result.error.type}`, }, ], isError: true as const, }; } const methodology = result.value; return { content: [ { type: "text" as const, text: JSON.stringify(methodology, null, 2), }, ], structuredContent: methodology as unknown as { [key: string]: unknown }, }; } ); // Tool: Audit model references against HUMMBL Base120 and methodology expectations server.registerTool( "audit_model_references", { title: "Audit HUMMBL Model References", description: "Audit a list of HUMMBL model references for existence, transformation alignment, and duplicates.", inputSchema: z.object({ items: z .array( z.object({ code: z.string().min(2).describe("Model code, e.g., IN11, CO4"), expectedTransformation: z .enum(["P", "IN", "CO", "DE", "RE", "SY"]) .optional() .describe("Optional expected transformation key for this reference"), }) ) .min(1), }), outputSchema: z.object({ methodologyId: z.string(), documentVersion: z.string(), totalReferences: z.number(), validCount: z.number(), invalidCount: z.number(), issues: z.array( z.object({ code: z.string(), issueType: z.enum(["NotFound", "WrongTransformation", "Duplicate", "Unknown"]), message: z.string(), expectedTransformation: z.string().optional(), actualTransformation: z.string().optional(), }) ), }), }, async ({ items }) => { const auditItems: AuditInputItem[] = items.map((item) => ({ code: item.code, expectedTransformation: item.expectedTransformation, })); const result = auditModelCodes(auditItems); if (!isOk(result)) { const error = result.error; const baseMessage = error.type === "ValidationError" ? error.message : `Unable to audit model references: ${error.type}`; return { content: [ { type: "text" as const, text: baseMessage, }, ], isError: true as const, }; } const payload = result.value; return { content: [ { type: "text" as const, text: JSON.stringify(payload, null, 2), }, ], structuredContent: payload as unknown as { [key: string]: unknown }, }; } ); }