Multi-Agent Software Development System for VS Code 🚀
An autonomous 5-agent AI pair-programming assistant integrated natively into VS Code. Designed for full-stack software development, architectural planning, automatic framework skill execution, and test verification.
🏗️ Architecture & Multi-Agent Workflow
flowchart TD
User["👤 Developer Prompt"] --> Agent5["🤖 Agent 5 (Project Manager)"]
Agent5 --> Plan["📄 Interactive Plan Card (Approve/Reject)"]
Plan -->|Approved| Delegation["⚡ Task Delegation"]
Delegation --> Agent1["🎨 Agent 1 (Frontend Developer)"]
Delegation --> Agent2["⚙️ Agent 2 (Backend Developer)"]
Delegation --> Agent3["🗄️ Agent 3 (Data Manager)"]
Agent1 & Agent2 & Agent3 --> Agent4["🧪 Agent 4 (QA & Test Supervisor)"]
Agent4 --> DiffCard["📊 Code Changes Diff Card"]
🌟 Key Features
Autonomous 5-Agent Team:
- 🤖 Agent 5 (Workspace Manager): Manages architectural state, plans tasks, updates
PROJECT_STATE.md, and delegates implementation.
- 🎨 Agent 1 (Frontend Developer): React, Vue, TSX/JSX layout, and UI component engineering.
- ⚙️ Agent 2 (Backend Developer): FastAPI REST endpoints, controllers, Pydantic schemas, and routing.
- 🗄️ Agent 3 (Data Manager): SQL schemas, database models, and Alembic migrations.
- 🧪 Agent 4 (Supervisor / QA): Pytest & Jest test suite generation and automated test verification.
Interactive Implementation Plans:
- Generates interactive plan cards with
[📄 Review Plan], [✅ Approve & Execute], and [❌ Reject / Edit] controls directly inside VS Code.
Unified Code Changes Card Widget:
- Tracks diffs across workspace files and displays a Cursor-style diff summary with direct review links to
CHANGES_LOG.md.
Multi-LLM Provider Support:
- Seamlessly switch between local Ollama (
qwen3-coder:latest), OpenAI (gpt-4o), Groq (llama-3.3-70b), DeepSeek, and Anthropic via VS Code Settings.
Automatic Skill Activation:
- Auto-activates built-in framework skills (
fastapi-backend, react-frontend, vue-frontend, database-migration, unit-testing) based on prompt context.
⚙️ Extension Settings
Access settings via Ctrl + , (or Cmd + ,) and search for Multi-Agent:
| Setting |
Type |
Default |
Description |
multiAgent.provider |
enum |
"ollama" |
LLM provider backend ("ollama", "openai", "groq", "deepseek", "anthropic"). |
multiAgent.apiKey |
string |
"" |
API key for cloud LLM providers. |
multiAgent.modelName |
string |
"qwen3-coder:latest" |
Default model name (e.g. 'qwen3-coder:latest', 'gpt-4o'). |
multiAgent.pythonPath |
string |
"python" |
Path to Python binary (or workspace .venv path). |
multiAgent.serverPort |
number |
8000 |
FastAPI server port number. |
multiAgent.serverHost |
string |
"127.0.0.1" |
FastAPI server host IP. |
multiAgent.autoStartServer |
boolean |
true |
Automatically launch backend server process on VS Code startup. |
multiAgent.requirePermission |
boolean |
true |
Require user approval before executing file writes or system commands. |
💻 Commands
Access from the Command Palette (Ctrl + Shift + P / Cmd + Shift + P):
Multi-Agent: Start Backend Server: Launches the Python backend server process manually.Multi-Agent: Restart Backend Server: Kills and restarts the server process.Multi-Agent: Open PROJECT_STATE.md: Opens persistent project memory in the editor.
🚀 Quickstart Guide
1. Install Extension
Install from the VS Code Extensions tab (Ctrl + Shift + X $\rightarrow$ search Multi-Agent Dev System).
- For Local Ollama: Ensure Ollama is running (
ollama run qwen3-coder:latest).
- For OpenAI / Cloud LLMs: Open Settings $\rightarrow$ set
multiAgent.provider to openai and paste your multiAgent.apiKey.
3. Open Workspace & Start Coding
Open any project folder, click the Multi-Agent System icon in the activity bar, and ask Agent 5 to build your feature!
| |
