<div align="center">

Awesome AI Tools

An open-source platform for discovering, managing, and deploying AI skills, subagents, and multi-agent workflows.

Platform · Documentation · CLI Reference · API Reference · Self-Hosting

</div>

---

Overview

Awesome AI Tools is a full-stack platform that centralizes 2,582 AI skills, 68 specialist subagents, and 4 composite multi-agent workflows. It ships as both a web application and a CLI tool.

Web platform — Browse, search, and export AI configurations for your IDE or agent runtime. Includes a universal AI chat interface with a 5-tier provider fallback (Groq, OpenRouter, DeepSeek, LLM7, and local synthesis).

CLI tool — Install skills directly into your workspace from the terminal. Compatible with Cursor, Claude Code, Windsurf, Cline, Roo Code, and GitHub Copilot.

---

Contents

• Platform
• CLI Reference
• Documentation
  • Quickstart
  • Skills System
  • Subagents
  • Multi-Agent Workflows
  • AI Chat & Provider Cascade
  • Local Proxy Router
  • IDE Config Export
• API Reference
• Self-Hosting
• Contributing

---

Platform

Live at https://awesome-ai-tools-suite.vercel.app

| Page | Path | Description | |:-----|:-----|:------------| | Home | / | Overview and ecosystem metrics | | Skills Browser | /skills | Browse and search all 2,582 AI skills | | Subagents | /agents | 68 specialist subagents directory | | Router | /router | Multi-agent workflow visualizer and local proxy setup | | AI Chat | /chat | Universal AI chat with slash command palette | | Stack Builder | /stack | Compose and export your agent stack | | Docs | /docs | Full documentation | | APIs | /apis | Public AI API directory |

---

CLI Reference

Installation

# Run without installing
npx awesome-ai-tools <command>

# Install globally
npm install -g awesome-ai-tools

Requirements: Node.js >= 18.0.0

Commands

init

Initialize your workspace with skills, subagents, hooks, and instincts memory.

npx awesome-ai-tools init

This creates:

• .agents/skills/ — all 2,582 skill files
• .agents/subagents/ — all 68 subagent definitions
• .claude/hooks/ — hook scripts for agent lifecycle events
• instincts.md — persistent codebase memory file

Options:

npx awesome-ai-tools init --minimal     # Skills only, no hooks
npx awesome-ai-tools init --dry-run     # Preview without writing files

---

pull

Download a domain-specific skill bundle into your workspace.

npx awesome-ai-tools pull <category>

Available categories:

| Category | Skills | Description | |:---------|:-------|:------------| | frontend | 320+ | React, Next.js, Vue, Tailwind, accessibility | | backend | 280+ | Node.js, FastAPI, Django, Rails, Go, Rust | | security | 190+ | Penetration testing, SAST, OWASP, secrets | | devops | 240+ | Docker, Kubernetes, Terraform, CI/CD, AWS | | architecture | 210+ | DDD, microservices, event sourcing, CQRS | | all | 2,582 | Full bundle |

Example:

npx awesome-ai-tools pull frontend
# Downloads skills to .agents/skills/frontend/

---

list

List all available skills, subagents, and workflows.

npx awesome-ai-tools list

---

scan

Run AgentShield security audit on your current workspace.

npx awesome-ai-tools scan
npx awesome-ai-tools scan --fix          # Apply auto-fixes where safe
npx awesome-ai-tools scan --report json  # Output as JSON

Checks for:

• Hardcoded secrets and API keys in source files
• Insecure environment variable usage
• Prompt injection vulnerabilities in agent instructions
• Overly permissive agent tool scopes

---

learn

Persist a rule or preference to instincts.md for agent memory.

npx awesome-ai-tools learn "Always use Zod for runtime validation"
npx awesome-ai-tools learn "Prefer server components over client components in Next.js"

Rules are appended to instincts.md and read by agents on every session start.

---

Documentation

Quickstart

Option A — Use the hosted platform

Visit https://awesome-ai-tools-suite.vercel.app. No account required.

Option B — CLI in an existing project

cd your-project
npx awesome-ai-tools init

Option C — Self-host

See Self-Hosting.

---

Skills System

A skill is a structured markdown instruction file that configures an AI agent's behavior for a specific domain. Each skill contains:

• A YAML frontmatter block with name and description
• Detailed instructions, patterns, and examples
• Optional reference files and scripts

Skills are stored in .agents/skills/<slug>/SKILL.md and are loaded automatically by compatible runtimes (Claude Code, Cursor, Windsurf, Cline).

Skill categories:

frontend/        backend/         security/
devops/          architecture/    mobile/
database/        testing/         ai-ml/
cloud/           design/          workflow/

Using a skill in Cursor:

npx awesome-ai-tools pull frontend
# Then in Cursor, reference .agents/skills/react-patterns/SKILL.md

Using a skill in Claude Code:

Skills are auto-loaded from .agents/skills/ when you run claude in a workspace initialized with npx awesome-ai-tools init.

---

Subagents

Subagents are specialist agent definitions configured for a narrow domain. Unlike general-purpose AI, each subagent has a fixed role, restricted tool scope, and domain-specific system prompt.

The 68 included subagents cover:

| Domain | Subagents | |:-------|:----------| | Frontend | nextjs-frontend-lead, tailwind-design-system-architect | | Backend | backend-microservices-architect, fastapi-async-engineer | | Security | security-auditor, prompt-red-teamer, dependency-vulnerability-scanner | | DevOps | devops-engineer, kubernetes-operator-dev, cicd-accelerator | | Data | python-data-engineer, clickhouse-olap-architect, duckdb-analytics-lead | | AI/ML | mlops-rag-engineer, prompt-engineer, vector-db-engineer | | Testing | playwright-e2e-architect, vitest-unit-master, tdd-driver | | Architecture | software-architect, clean-architecture-modeler, event-sourcing-cqrs-architect |

Subagent definitions are in .agents/subagents/<name>.md.

---

Multi-Agent Workflows

Four composite workflows are available via the /router page and the trigger command.

/review — Code & Security Review

Two-stage review pipeline:

1. AgentShield scans for security vulnerabilities, injections, and exposed secrets
2. Code Logic Reviewer checks correctness, race conditions, and standards compliance

# Trigger from CLI (coming in v1.1)
npx awesome-ai-tools trigger review

/tdd — Autonomous TDD Cycle

Strict Red-Green-Refactor:

1. Red — writes failing unit tests against the spec
2. Green — implements minimal passing code
3. Refactor — safely refactors with test coverage protection

/council — Multi-Model Deliberation

Routes an architectural decision through multiple reasoning perspectives before writing implementation code. Use before major system design choices.

/compact — Context Compaction

Summarizes active conversation history, clears redundant tokens, and extracts key codebase instincts. Run when context window is degrading.

---

AI Chat & Provider Cascade

The chat interface at /chat routes requests through a 5-tier fallback cascade:

Tier 1  BYOK           User-supplied API key via Settings modal
Tier 2  Server keys    GROQ → OPENROUTER → DEEPSEEK → OPENAI
Tier 3  Ollama         http://localhost:11434 (self-hosted only)
Tier 4  LLM7           Free DeepSeek-V4-Flash (no key required)
Tier 5  Local          Contextual knowledge synthesis (zero downtime)

The AI responds regardless of API key configuration. Tiers 1–2 provide the best quality and speed.

Slash command palette:

Type / in the chat input to open the command palette. All 2,582 skills and 4 workflows are searchable and insertable inline.

---

Local Proxy Router

The router at /router exposes a local OpenAI-compatible API endpoint. Point any IDE or client to it instead of calling providers directly.

Setup:

git clone https://github.com/dimasrahmandaalfarizi/Awesome-AI-Tools
cd Awesome-AI-Tools
npm install
npm run dev

Endpoint: http://localhost:3000/api/v1

Configure your IDE:

| IDE | Setting | Value | |:----|:--------|:------| | Cursor | OpenAI Base URL | http://localhost:3000/api/v1 | | Cline | API Base URL | http://localhost:3000/api/v1 | | Roo Code | Custom API Base | http://localhost:3000/api/v1 | | Continue | apiBase | http://localhost:3000/api/v1 |

Leave the API key field blank or enter any string. The router forwards requests using your configured provider credentials.

Supported providers:

• Groq (recommended — fast, free tier)
• OpenRouter
• DeepSeek
• Google Gemini
• OpenAI
• Ollama (local)

---

IDE Config Export

From /stack or /chat, click Export IDE Config to generate ready-to-use configuration files:

| IDE | Output file | |:----|:-----------| | Cursor | .cursorrules | | VS Code Continue | .continue/config.json | | Claude Desktop | claude_desktop_config.json | | Windsurf | .windsurfrules |

---

API Reference

Base URL (production): https://awesome-ai-tools-suite.vercel.app Base URL (local): http://localhost:3000

POST /api/chat/universal

Universal chat endpoint with provider cascade.

Request:

{
  "messages": [
    { "role": "user", "content": "Explain server components in Next.js" }
  ],
  "provider": "groq",
  "model": "llama-3.3-70b-versatile",
  "apiKey": "opti…[redacted]"
}

Response: Server-sent events (SSE) stream.

data: {"content": "Server components are..."}
data: {"content": " rendered on the server"}
data: [DONE]

---

POST /api/v1/chat/completions

OpenAI-compatible endpoint. Drop-in replacement for api.openai.com/v1/chat/completions.

Request:

{
  "model": "gpt-4o",
  "messages": [{ "role": "user", "content": "Hello" }],
  "stream": true
}

---

GET /api/search

Full-text search across skills, subagents, tools, and APIs.

GET /api/search?q=react+hooks&type=skills

Parameters:

| Param | Type | Description | |:------|:-----|:------------| | q | string | Search query | | type | string | skills, agents, tools, apis, or all | | limit | number | Max results (default: 20) |

---

GET /api/ai/semantic-search

Semantic vector search across the skills database.

GET /api/ai/semantic-search?q=rate+limiting+middleware&limit=5

---

GET /api/router/config

Returns the current proxy router configuration.

---

Self-Hosting

Vercel (Recommended)

Set the following enviro