AgentSys

Overview

Imagine an AI assistant that doesn’t just write code — it manages your entire development workflow. It plans features, writes tests, refactors legacy systems, deploys updates, and even monitors performance — all while you focus on high-level design and innovation.

That future is here, and it’s called AgentSys.

AgentSys is not just another AI tool — it’s a full-fledged AI automation system engineered to transform how developers interact with code. With 26 powerful plugins, 50 specialized AI agents, and 47 ready-to-deploy skills, AgentSys turns your IDE into a self-driving development environment. Whether you're building a simple script or managing a complex microservices architecture, AgentSys automates repetitive tasks, accelerates delivery, and reduces cognitive load.

Designed from the ground up for multi-platform compatibility, AgentSys integrates seamlessly with leading AI coding environments including Claude Code, OpenCode, Codex, Cursor, and Kiro. And the best part? It's completely free and open-source.

In today’s fast-paced software landscape, manual coding isn’t scalable. The real competitive edge goes to teams who can automate intelligently — and AgentSys delivers exactly that: smart, modular, and extensible automation at scale.

Whether you’re a solo developer looking to boost productivity or part of a team aiming to standardize workflows, AgentSys empowers you to work faster, smarter, and with fewer errors.

Let’s dive into what makes this system so powerful.

Key Benefits

1. End-to-End Workflow Automation

AgentSys eliminates context switching by handling everything from idea to deployment. For example:

You describe a new feature in natural language.
An agent parses the request, generates user stories, creates a task breakdown, writes unit tests, implements logic, runs CI checks, and opens a PR.
All without leaving your editor.

👉 Ideal for sprint planning, rapid prototyping, or onboarding new developers.

2. Modular Architecture with Plug-and-Play Agents

Each of the 50 agents specializes in a specific function — like TestGeneratorAgent, BugHunterAgent, or DocWriterAgent. These can be chained together or used independently based on your needs.

Need only documentation help? Use the DocWriterAgent alone.
Building a full-stack module? Chain APIPlannerAgent → FrontendBuilderAgent → BackendEngineerAgent.

👉 Perfect for customizing workflows without reinventing the wheel.

3. Cross-Platform Compatibility

Unlike most AI tools locked into one ecosystem, AgentSys supports Claude Code, Cursor, Codex, OpenCode, and Kiro, making it ideal for hybrid teams using different tools.

No more vendor lock-in. Your automation travels with you.

👉 Great for organizations standardizing processes across diverse tech stacks.

4. Accelerated Debugging & Code Refactoring

The built-in BugHunterAgent analyzes error logs, correlates them with recent commits, reproduces issues in sandboxed environments, and suggests fixes — often before you even notice the bug.

Meanwhile, RefactorProAgent identifies code smells, technical debt, and performance bottlenecks, then rewrites modules safely using AST-based transformations.

👉 Reduces debugging time by up to 70% in real-world projects.

5. Instant Onboarding & Knowledge Transfer

New team members get instant access to institutional knowledge through OnboarderAgent, which generates contextual guides, explains architecture decisions, and walks through key components.

Even better: every agent learns from past interactions (locally, no data leakage), so your system gets smarter over time.

👉 Cuts ramp-up time from weeks to hours.

Core Features

Feature	Description	Supported Platforms
26 Plugins	Extend functionality with domain-specific plugins (e.g., REST API scaffolding, database migration, security auditing)	All platforms
50 Specialized Agents	Pre-trained AI agents for testing, documentation, debugging, deployment, etc.	All platforms
47 Ready-to-Use Skills	Reusable automation patterns like “Create CRUD Endpoint” or “Migrate Legacy Module”	All platforms
Natural Language Planning Engine	Describe tasks in plain English; AgentSys breaks them into executable steps	Claude Code, Cursor, Kiro
Multi-Agent Collaboration	Agents communicate internally to coordinate complex workflows (e.g., frontend + backend sync)	All platforms
Local Execution Mode	Run agents entirely offline for sensitive codebases (enterprise-grade privacy)	Cursor, OpenCode, Kiro
GitHub/GitLab Integration	Auto-create branches, commit changes, open pull requests with review summaries	All platforms
Custom Rule Configuration	Define rules for code style, security policies, or approval gates via `.cursorrules` or JSON config	Cursor, Codex

How to Get & Install

AgentSys is designed for universal adoption. While some platforms offer deeper integration than others, you can install and use AgentSys regardless of your preferred IDE or AI assistant.

Here’s how to get started — step by step:

✅ Universal Installation (All Platforms)

Since AgentSys is hosted on GitHub and built as a modular framework, the primary installation method is direct integration via source.

Step 1: Clone the Repository

Open your terminal and run:

git clone https://github.com/agent-sh/agentsys.git
cd agentsys

Step 2: Install Dependencies

Ensure Python 3.9+ is installed, then:

pip install -r requirements.txt

Note: Some agents require additional dependencies (e.g., Docker for sandboxing). See docs/setup.md for advanced configurations.

Step 3: Initialize the System

Run the setup wizard:

python main.py --init

This will guide you through:

Selecting your platform (Claude Code, Cursor, etc.)
Configuring API keys (if needed)
Choosing default agents and skills
Setting up local execution preferences

You’ll end up with a personalized agentsys-config.json file.

Step 4: Activate in Your Editor

Now connect AgentSys to your coding environment.

🔧 For Cursor Users

Cursor natively supports rule-based automation. To enable AgentSys:

Option A: Configure `.cursorrules`

Create a .cursorrules file in your project root:

{
  "plugins": [
    "./agentsys/plugins/rest-scaffold-plugin",
    "./agentsys/plugins/db-migration-tool"
  ],
  "agents": [
    "TestGeneratorAgent",
    "BugHunterAgent"
  ],
  "skills": [
    "create-crud-endpoint",
    "refactor-module"
  ],
  "model": "claude-3-opus",
  "temperature": 0.3
}

Once saved, restart Cursor. You’ll see AgentSys suggestions appear during coding sessions.

Option B: Use Built-In Command Palette

In Cursor, press Ctrl/Cmd + Shift + P, type /plugin install, and enter:

/plugin install https://github.com/agent-sh/agentsys

Follow prompts to complete installation.

🤖 For Claude Code / OpenCode Users

Claude Code supports third-party plugins through its marketplace.

Step 1: Open Plugin Marketplace

In your Claude interface, go to:

Settings → Integrations → Plugin Marketplace

Step 2: Search for AgentSys

Search for AgentSys and click Install.

If not yet listed, use manual plugin registration:

Go to: Custom Plugins > Register Remote Plugin

Enter URL: https://raw.githubusercontent.com/agent-sh/agentsys/main/plugin-manifest.json

Step 3: Enable Desired Agents

After installation, use slash commands to activate agents:

/agent enable TestGeneratorAgent
/skill apply create-unit-tests

You can now say:

“Write tests for this function” — and AgentSys handles the rest.

💻 For Codex & Kiro Users

These platforms support external skill loading via configuration files.

In Your Project Root, Create `skills.json`:

{
  "skills": [
    "tdd-workflow",
    "api-scaffolder",
    "security-audit"
  ],
  "agents": [
    "OnboarderAgent",
    "RefactorProAgent"
  ],
  "autostart": true
}

Then launch Kiro or Codex within the directory — it will auto-detect and load AgentSys components.

Alternatively, import individual skills directly from the GitHub releases page.

🎉 You’re Now Ready!

Launch any supported editor, describe a task like:

“Build a login API with JWT auth and rate limiting”

…and watch AgentSys plan, code, test, and document the solution — autonomously.

Use Cases

1. Rapid Prototyping for Startups

Founders can go from concept to MVP in hours instead of days. Just describe the app flow, and let APICreatorAgent + FrontendScaffolderAgent generate working endpoints and UI components.

💡 Save $10k+ in dev costs during early-stage validation.

2. Enterprise Code Modernization

Legacy Java monoliths slowing you down? Deploy LegacyMigrationAgent to analyze old code, extract business logic, and rebuild services in modern TypeScript with full test coverage.

🔁 Automate tech debt reduction without risking downtime.

3. CI/CD Pipeline Enhancement

Integrate AgentSys into your GitHub Actions pipeline. When a PR is opened:

CodeReviewerAgent analyzes changes
TestCoverageAgent ensures minimum thresholds
SecurityScannerAgent flags vulnerabilities
Results are posted as comments

🛡️ Strengthen quality gates with AI precision.

4. Developer Onboarding Acceleration

Assign new hires a starter ticket like:

“Fix a typo in the dashboard”

With AgentSys enabled, they receive interactive guidance:

Where the file lives
How to run the dev server
What tests to check
How to submit a PR

🚀 Turn first-commit anxiety into confidence.

5. Documentation That Stays Updated

Forget outdated wikis. DocWriterAgent watches code changes and auto-generates or updates:

API references
Architecture diagrams (via Mermaid.js)
Changelog entries
READMEs

📚 Keep docs in sync with zero manual effort.

Tips for Getting the Most Out of AgentSys

Start Small, Then Scale Up
Begin with one agent (e.g., TestGeneratorAgent) before enabling full multi-agent orchestration. This helps you understand behavior patterns and build trust.
Customize Agent Chains
Combine agents creatively. Example:
```
PlanFeatureAgent → WriteTestsAgent → ImplementCodeAgent → ReviewPRAgent
```
Save these chains as reusable templates in workflows/.
Use Local Mode for Sensitive Projects
Set "execution": "local" in your config to ensure all processing happens on-machine — ideal for financial, healthcare, or government-related code.

Disclaimer: AgentSys is an open-source project developed by the community. While extensive testing has been performed, always review automated changes before merging into production. The authors assume no liability for direct or consequential damages arising from misuse or unexpected behavior. Use responsibly and maintain human oversight in critical systems.

Overview

That future is here, and it’s called AgentSys.

Whether you’re a solo developer looking to boost productivity or part of a team aiming to standardize workflows, AgentSys empowers you to work faster, smarter, and with fewer errors.

Let’s dive into what makes this system so powerful.

Key Benefits

1. End-to-End Workflow Automation

AgentSys eliminates context switching by handling everything from idea to deployment. For example:

You describe a new feature in natural language.
An agent parses the request, generates user stories, creates a task breakdown, writes unit tests, implements logic, runs CI checks, and opens a PR.
All without leaving your editor.

👉 Ideal for sprint planning, rapid prototyping, or onboarding new developers.

2. Modular Architecture with Plug-and-Play Agents

Each of the 50 agents specializes in a specific function — like TestGeneratorAgent, BugHunterAgent, or DocWriterAgent. These can be chained together or used independently based on your needs.

Need only documentation help? Use the DocWriterAgent alone.
Building a full-stack module? Chain APIPlannerAgent → FrontendBuilderAgent → BackendEngineerAgent.

👉 Perfect for customizing workflows without reinventing the wheel.

3. Cross-Platform Compatibility

Unlike most AI tools locked into one ecosystem, AgentSys supports Claude Code, Cursor, Codex, OpenCode, and Kiro, making it ideal for hybrid teams using different tools.

No more vendor lock-in. Your automation travels with you.

👉 Great for organizations standardizing processes across diverse tech stacks.

4. Accelerated Debugging & Code Refactoring

The built-in BugHunterAgent analyzes error logs, correlates them with recent commits, reproduces issues in sandboxed environments, and suggests fixes — often before you even notice the bug.

Meanwhile, RefactorProAgent identifies code smells, technical debt, and performance bottlenecks, then rewrites modules safely using AST-based transformations.

👉 Reduces debugging time by up to 70% in real-world projects.

5. Instant Onboarding & Knowledge Transfer

New team members get instant access to institutional knowledge through OnboarderAgent, which generates contextual guides, explains architecture decisions, and walks through key components.

Even better: every agent learns from past interactions (locally, no data leakage), so your system gets smarter over time.

👉 Cuts ramp-up time from weeks to hours.

Core Features

Feature	Description	Supported Platforms
26 Plugins	Extend functionality with domain-specific plugins (e.g., REST API scaffolding, database migration, security auditing)	All platforms
50 Specialized Agents	Pre-trained AI agents for testing, documentation, debugging, deployment, etc.	All platforms
47 Ready-to-Use Skills	Reusable automation patterns like “Create CRUD Endpoint” or “Migrate Legacy Module”	All platforms
Natural Language Planning Engine	Describe tasks in plain English; AgentSys breaks them into executable steps	Claude Code, Cursor, Kiro
Multi-Agent Collaboration	Agents communicate internally to coordinate complex workflows (e.g., frontend + backend sync)	All platforms
Local Execution Mode	Run agents entirely offline for sensitive codebases (enterprise-grade privacy)	Cursor, OpenCode, Kiro
GitHub/GitLab Integration	Auto-create branches, commit changes, open pull requests with review summaries	All platforms
Custom Rule Configuration	Define rules for code style, security policies, or approval gates via `.cursorrules` or JSON config	Cursor, Codex

How to Get & Install

AgentSys is designed for universal adoption. While some platforms offer deeper integration than others, you can install and use AgentSys regardless of your preferred IDE or AI assistant.

Here’s how to get started — step by step:

✅ Universal Installation (All Platforms)

Since AgentSys is hosted on GitHub and built as a modular framework, the primary installation method is direct integration via source.

Step 1: Clone the Repository

Open your terminal and run:

git clone https://github.com/agent-sh/agentsys.git
cd agentsys

Step 2: Install Dependencies

Ensure Python 3.9+ is installed, then:

pip install -r requirements.txt

Note: Some agents require additional dependencies (e.g., Docker for sandboxing). See docs/setup.md for advanced configurations.

Step 3: Initialize the System

Run the setup wizard:

python main.py --init

This will guide you through:

Selecting your platform (Claude Code, Cursor, etc.)
Configuring API keys (if needed)
Choosing default agents and skills
Setting up local execution preferences

You’ll end up with a personalized agentsys-config.json file.

Step 4: Activate in Your Editor

Now connect AgentSys to your coding environment.

🔧 For Cursor Users

Cursor natively supports rule-based automation. To enable AgentSys:

Option A: Configure `.cursorrules`

Create a .cursorrules file in your project root:

{
  "plugins": [
    "./agentsys/plugins/rest-scaffold-plugin",
    "./agentsys/plugins/db-migration-tool"
  ],
  "agents": [
    "TestGeneratorAgent",
    "BugHunterAgent"
  ],
  "skills": [
    "create-crud-endpoint",
    "refactor-module"
  ],
  "model": "claude-3-opus",
  "temperature": 0.3
}

Once saved, restart Cursor. You’ll see AgentSys suggestions appear during coding sessions.

Option B: Use Built-In Command Palette

In Cursor, press Ctrl/Cmd + Shift + P, type /plugin install, and enter:

/plugin install https://github.com/agent-sh/agentsys

Follow prompts to complete installation.

🤖 For Claude Code / OpenCode Users

Claude Code supports third-party plugins through its marketplace.

Step 1: Open Plugin Marketplace

In your Claude interface, go to:

Settings → Integrations → Plugin Marketplace

Step 2: Search for AgentSys

Search for AgentSys and click Install.

If not yet listed, use manual plugin registration:

Go to: Custom Plugins > Register Remote Plugin

Enter URL: https://raw.githubusercontent.com/agent-sh/agentsys/main/plugin-manifest.json

Step 3: Enable Desired Agents

After installation, use slash commands to activate agents:

/agent enable TestGeneratorAgent
/skill apply create-unit-tests

You can now say:

“Write tests for this function” — and AgentSys handles the rest.

💻 For Codex & Kiro Users

These platforms support external skill loading via configuration files.

In Your Project Root, Create `skills.json`:

{
  "skills": [
    "tdd-workflow",
    "api-scaffolder",
    "security-audit"
  ],
  "agents": [
    "OnboarderAgent",
    "RefactorProAgent"
  ],
  "autostart": true
}

Then launch Kiro or Codex within the directory — it will auto-detect and load AgentSys components.

Alternatively, import individual skills directly from the GitHub releases page.

🎉 You’re Now Ready!

Launch any supported editor, describe a task like:

“Build a login API with JWT auth and rate limiting”

…and watch AgentSys plan, code, test, and document the solution — autonomously.

Use Cases

1. Rapid Prototyping for Startups

Founders can go from concept to MVP in hours instead of days. Just describe the app flow, and let APICreatorAgent + FrontendScaffolderAgent generate working endpoints and UI components.

💡 Save $10k+ in dev costs during early-stage validation.

2. Enterprise Code Modernization

Legacy Java monoliths slowing you down? Deploy LegacyMigrationAgent to analyze old code, extract business logic, and rebuild services in modern TypeScript with full test coverage.

🔁 Automate tech debt reduction without risking downtime.

3. CI/CD Pipeline Enhancement

Integrate AgentSys into your GitHub Actions pipeline. When a PR is opened:

CodeReviewerAgent analyzes changes
TestCoverageAgent ensures minimum thresholds
SecurityScannerAgent flags vulnerabilities
Results are posted as comments

🛡️ Strengthen quality gates with AI precision.

4. Developer Onboarding Acceleration

Assign new hires a starter ticket like:

“Fix a typo in the dashboard”

With AgentSys enabled, they receive interactive guidance:

Where the file lives
How to run the dev server
What tests to check
How to submit a PR

🚀 Turn first-commit anxiety into confidence.

5. Documentation That Stays Updated

Forget outdated wikis. DocWriterAgent watches code changes and auto-generates or updates:

API references
Architecture diagrams (via Mermaid.js)
Changelog entries
READMEs

📚 Keep docs in sync with zero manual effort.

Tips for Getting the Most Out of AgentSys

Start Small, Then Scale Up
Begin with one agent (e.g., TestGeneratorAgent) before enabling full multi-agent orchestration. This helps you understand behavior patterns and build trust.
Customize Agent Chains
Combine agents creatively. Example:
```
PlanFeatureAgent → WriteTestsAgent → ImplementCodeAgent → ReviewPRAgent
```
Save these chains as reusable templates in workflows/.
Use Local Mode for Sensitive Projects
Set "execution": "local" in your config to ensure all processing happens on-machine — ideal for financial, healthcare, or government-related code.

[AI Skill] AgentSys: Features & Installation Guide

Overview

Key Benefits

1. End-to-End Workflow Automation

2. Modular Architecture with Plug-and-Play Agents

3. Cross-Platform Compatibility

4. Accelerated Debugging & Code Refactoring

5. Instant Onboarding & Knowledge Transfer

Core Features

How to Get & Install

✅ Universal Installation (All Platforms)

Step 1: Clone the Repository

Step 2: Install Dependencies

Step 3: Initialize the System

Step 4: Activate in Your Editor

🔧 For Cursor Users

Option A: Configure .cursorrules

Option B: Use Built-In Command Palette

🤖 For Claude Code / OpenCode Users

Step 1: Open Plugin Marketplace

Step 2: Search for AgentSys

Step 3: Enable Desired Agents

💻 For Codex & Kiro Users

In Your Project Root, Create skills.json:

Use Cases

1. Rapid Prototyping for Startups

2. Enterprise Code Modernization

3. CI/CD Pipeline Enhancement

4. Developer Onboarding Acceleration

5. Documentation That Stays Updated

Tips for Getting the Most Out of AgentSys

Related Skills

CI/CD Pipeline Generator

Shell Script Automation

Production-Grade Engineering Skills

AgentSys

[AI Skill] AgentSys: Features & Installation Guide

Overview

Key Benefits

1. End-to-End Workflow Automation

2. Modular Architecture with Plug-and-Play Agents

3. Cross-Platform Compatibility

4. Accelerated Debugging & Code Refactoring

5. Instant Onboarding & Knowledge Transfer

Core Features

How to Get & Install

✅ Universal Installation (All Platforms)

Step 1: Clone the Repository

Step 2: Install Dependencies

Step 3: Initialize the System

Step 4: Activate in Your Editor

🔧 For Cursor Users

Option A: Configure .cursorrules

Option B: Use Built-In Command Palette

🤖 For Claude Code / OpenCode Users

Step 1: Open Plugin Marketplace

Step 2: Search for AgentSys

Step 3: Enable Desired Agents

💻 For Codex & Kiro Users

In Your Project Root, Create skills.json:

Use Cases

1. Rapid Prototyping for Startups

2. Enterprise Code Modernization

3. CI/CD Pipeline Enhancement

4. Developer Onboarding Acceleration

5. Documentation That Stays Updated

Tips for Getting the Most Out of AgentSys

Related Skills

CI/CD Pipeline Generator

Shell Script Automation

Production-Grade Engineering Skills

Option A: Configure `.cursorrules`

In Your Project Root, Create `skills.json`:

Option A: Configure `.cursorrules`

In Your Project Root, Create `skills.json`: