SWESim - Software Engineering Simulation Network

A powerful multi-agent system that coordinates specialized AI agents to generate comprehensive software implementation decision reports for full-stack applications.

Usecase

This project aims to solve the problem of brainstorming how an full stack app would look like just from having an idea, this project helps developers explore ideas agents present to them via decision reports and change agents according to stack they use in development. It enables complete software project planning through coordinated AI agents using the Mastra framework. It provides automated design, architecture, and implementation planning for web applications, all orchestrated through a unified agent network.

Template

This template provides easy to use agentnetwork that uses Mastra's tools, agents and agentnetwork that can be forked for personal needs, ex: changing backend from Typescript to Java or making Database agent use NoSQL instead of SQL

Features

• Multi-Agent Coordination: Four specialized agents working in sequence
• Complete Project Planning: From database design to UI/UX specifications
• Implementation Reports: Detailed technical specifications saved locally
• Sequential Workflow: Agents build upon each other's work for cohesive results
• Modern Tech Stack: Built with Mastra framework and OpenAI integration

Installation

Prerequisites

• Node.js (v20+)
• npm or yarn
• OpenAI API access

Setup

1. Clone the repository:

git clone https://github.com/jalpp/swesim.git
cd swesim

2. Install dependencies:

npm install

3. Copy example env file

  cp .env.example .env

4. Insert the OpenAI API key

OPENAI_API_KEY=insert-key-here

5. Start the Mastra server

 npm run dev

Usage

Head over to Networks tab -> select SWE Simulation Network -> prompt it to build a project idea

Generated Reports

Implementation reports are automatically saved to ./mastra/reports/ folder with complete technical specifications.

Example Reports

the /reports folder contains sample report which was generated when asked to create a clothing line shopping full stack app.

Architecture

Core Agents

1. Database Agent

   • Designs database schemas and data models
   • Creates migration scripts and indexing strategies
   • Provides performance optimization recommendations

2. Backend Agent

   • Creates API specifications and server architecture
   • Implements authentication and security patterns
   • Designs scalable backend infrastructure

3. Frontend Agent

   • Plans React/Next.js component architecture
   • Defines state management and data flow
   • Implements responsive design patterns

4. Designer Agent

   • Develops UI/UX specifications and design systems
   • Creates accessibility guidelines and user flows
   • Establishes visual design standards

Coordinator Network(AgentNetwork)

• Orchestrates all agents in optimal sequence
• Ensures context sharing between agents
• Compiles unified implementation reports

Flow Diagram

Project Request → SWE Network → Database Agent → Backend Agent → Frontend Agent → Designer Agent → Local report generation 

Project Structure

swesim/
├── .mastra/output/reports/       # Generated implementation reports
├── reports/       # Example reports
├── src/

├── mastra/
│   ├── agents/        # Agent definitions and configurations
│   └── network/       # Coordinator network implementation
│   └── tools/       # tools implementation

Credits

This project is built with:

• Mastra - AI Agent framework
• OpenAI - AI models for agent intelligence