Skip to content

An AI-enhanced development environment where intelligent agents assist in coding, refactoring, and debugging, all directly within your IDE.

Notifications You must be signed in to change notification settings

ahmedcharef/Agent-Powered-IDE

Repository files navigation

Agentic Hybrid Dev Workflow: Kiro + Cursor + MCP + GitHub

Project Name (working title): Hybrid-Agent E-Commerce Lab
Goal: Demonstrate a powerful, traceable multi-agent development loop where two leading agentic IDEs collaborate on the same codebase via shared GitHub issues, Model Context Protocol (MCP), and role-based task splitting.

Overview

This repository experiments with agentic software development at scale in 2026:

  • Kiro (AWS agentic IDE) handles structured specification generation, backend implementation (APIs, database, auth, business logic), and autonomous execution triggered by GitHub webhooks (via "kiro" label or mentions).
  • Cursor (agentic VS Code fork) handles fast, creative frontend implementation (UI components, pages, client-side logic, styling) using Agent/Composer modes.
  • GitHub serves as the single source of truth:
    • Tasks & specs live as issues (created automatically from Kiro specs)
    • Labels split work: backend, frontend, task, mvp, etc.
    • PRs link back to issues
    • Agents push commits, create PRs, add review comments, and close issues with summaries on completion
  • Model Context Protocol (MCP) enables both IDEs to securely read/write GitHub issues, push code, and share context (via official GitHub MCP server + optional orchestration servers). Cursor v2.4.x has solid MCP fixes for tool detection.

The result: A semi-autonomous, self-healing development loop with full traceability, minimal manual git thrashing, and clear separation of concerns between structured (Kiro) and exploratory (Cursor) agents.

Core Workflow

  1. Spec & Task Generation (Kiro)

    • Feed project idea / requirements into Kiro
    • Agent generates EARS-style spec + task decomposition
    • Automatically creates GitHub issues with labels (backend / frontend)
  2. Backend Execution (Kiro Autonomous Agent – preferred)

    • Add "kiro" label to backend issues → webhook triggers autonomous agent
    • Agent clones sandbox, implements, commits to branch, opens PR, adds details
    • On merge: Closes issue with comment (summary, tests, notes)
  3. Frontend Execution (Cursor)

    • Persistent chat / Agent mode watches for frontend-labeled issues
    • Implements UI autonomously, pushes to branch, creates PR
    • Closes issue post-merge with visual/UI notes
  4. Review & Close

    • Cross-review PRs in either IDE
    • Merge → agents auto-close linked issue with changelog

Tech Stack (Initial Test Project)

  • Framework: Next.js 14+ (App Router)

  • Base Starter: https://github.com/kleneway/next-ai-starter – Next.js enables you to create full-stack web applications by extending the latest React features, and integrating powerful Rust-based JavaScript tooling for the fastest builds.

  • Database: -

  • Auth: -

  • Testing: Vitest

  • MCP Servers:

    • Official GitHub MCP (issues, PRs, push_files, etc.)
    • Optional: agent-orchestration (shared memory, task queue, role claiming)
  • IDE Agents:

    • Kiro Autonomous Agent (webhook-driven backend)
    • Cursor Agent/Composer (chat-driven frontend)

Getting Started

Prerequisites

  • Docker (for local GitHub MCP server)
  • GitHub PAT with repo scope
  • Installed: Kiro IDE, Cursor (free Hobby tier)

Setup Steps

  1. Fork/clone this repo
  2. Install dependencies: npm install
  3. Configure MCP in both IDEs (see below)

MCP Configuration (GitHub Integration)

Kiro (~/.kiro/settings/mcp.json or workspace .kiro/settings/mcp.json):

{
  "mcpServers": {
    "github-mcp": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "-e", "GITHUB_PERSONAL_ACCESS_TOKEN", "-e", "GITHUB_TOOLSETS=repos,issues,pull_requests,git", "ghcr.io/github/github-mcp-server"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "${input:github_pat}" }
    }
  },
  "inputs": [{ "type": "promptString", "id": "github_pat", "description": "GitHub PAT", "password": true }]
}

About

An AI-enhanced development environment where intelligent agents assist in coding, refactoring, and debugging, all directly within your IDE.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published