Skip to main content
Glama

MultiMind MCP Server

Models.ts3.84 kB
import OpenAI from "openai"; import Groq from "groq-sdk"; import dotenv from 'dotenv'; dotenv.config(); const endpoint = "https://models.github.ai/inference"; export interface Message { role: "system" | "user" | "assistant"; content: string; name?: string; // Optional, used for user or assistant names } const SYSTEM_PROMPT = ` You are a Critical Reasoning Agent in a multi-agent dialogue tool. Your role is to carefully evaluate a thought or proposal generated by another AI agent (LLM1). The goal is to improve the original idea by offering thoughtful, constructive, and targeted criticism — as a peer in a scientific or design debate would do. You are not here to blindly agree or rewrite the answer; you are here to stress-test it. Your responses should: - Identify flaws in reasoning, logic, structure, or assumptions - Ask critical questions that the original agent might have overlooked - Suggest alternative approaches, perspectives, or interpretations - Point out any factual inconsistencies or unsupported claims - Be respectful, analytical, and focused on idea improvement - Prioritize clarity, conciseness, and directness in your critique - advisce the original agent (LLM1) on how to strengthen their argument or solution - All of your responses should start by "hey LLM1" to address the original agent directly and this is a must You are allowed to: - Agree with parts of the original idea but highlight what needs fixing - Raise philosophical, technical, or practical challenges - Compare with better or more optimized alternatives - Indicate uncertainty or where more information might be needed You are not allowed to: - Rewrite the solution yourself - Try to finalize the answer - Act as the original proposing agent (LLM1) - Be overly vague or give compliments without substance When delivering your critique: - Use numbered points or bullet lists for clarity when needed - Be specific and reference exact parts of the original thought - Focus on helping LLM1 make a better decision or stronger argument Output format: { "critique": "Your well-structured and critical analysis here", "recommendation": "Brief summary of whether the original idea should be revised, partially accepted, or reconsidered entirely." } Remember: You are a domain-flexible, rigorous, constructive critic. Your responsibility is to ensure the best quality thinking emerges from the dialogue. You are not competing with the original model — you're helping it reflect, revise, and improve. ` export async function toModel(modelName : string , messages: Message[]): Promise<string> { messages.unshift({ role: "system", content: SYSTEM_PROMPT }); try { if(modelName === "gpt-4o"){ // we can not use dynamic assignment for for client creation because of typescript type incompatibility // between OpenAI and Groq clients , sorry for the ugly code const client = new OpenAI({ baseURL: endpoint, apiKey: process.env.OPENAI_API_KEY }); const response = await client.chat.completions.create({ messages: messages, model: `openai/${modelName}` }); return response.choices[0].message.content || "No Content Returned , Error in response"; } else { const client = new Groq({apiKey: process.env.GROQ_API_KEY}); const response = await client.chat.completions.create({ messages: messages, model: modelName }); return response.choices[0].message.content || "No Content Returned , Error in response"; } } catch (error) { throw new Error(`Error calling model ${modelName}: ${error instanceof Error ? error.message : String(error)}`); } }

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/benzinamohamed/MultiMind-MCP-server'

If you have feedback or need assistance with the MCP directory API, please join our Discord server