# Ai Toolkit By Role > 🧰 Beginner-friendly, filterable directory of AI skills, connectors & plugins β€” organized by your job, across Claude, ChatGPT, Cursor, Codex & more. - **Type:** MCP server - **Install:** `agentstack add mcp-rahul-khare997-ai-toolkit-by-role` - **Verified:** Yes β€” security-reviewed for prompt injection and unsafe behavior - **Seller:** [Rahul-Khare997](https://agentstack.voostack.com/s/rahul-khare997) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [Rahul-Khare997](https://github.com/Rahul-Khare997) - **Source:** https://github.com/Rahul-Khare997/ai-toolkit-by-role ## Install ```sh agentstack add mcp-rahul-khare997-ai-toolkit-by-role ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # 🧰 AI Toolkit by Role **New to AI and not sure what to install? Start here.** A plain-English, filterable directory of the AI **skills**, **connectors**, and **plugins** worth setting up β€” organized by *your job*, mapped across Claude, ChatGPT, Cursor, Codex, Gemini and more, and ranked by real **GitHub stars**. ### πŸ‘‰ [**Open the filterable website**](https://rahul-khare997.github.io/ai-toolkit-by-role/) β€” search and filter by role, tool, type, and where it works. > ⭐ Star counts are pulled live from GitHub and refreshed weekly (as of 2026-07-13). 🧰 = built into Claude already, so you may not need to install it. --- ## πŸ€” First, what are these things? (60-second explainer) If you're new to AI coding and chatting, the words are confusing. Here they are in plain English: - **🧠 Skill** β€” A packaged set of instructions that teaches your AI a specialized task β€” like writing a legal privacy notice, building a slide deck, or cleaning a spreadsheet. You 'install' it and then just ask for the task. - **πŸ”Œ Connector (MCP)** β€” A bridge that lets your AI reach an outside app or data source β€” GitHub, Gmail, a database, Slack. Most use a standard called MCP (Model Context Protocol). Think of it as giving your AI a new sense or a new hand. - **🧩 Plugin** β€” A bundle you install into a tool that adds a set of features at once β€” often several skills, commands, and connectors packaged together. - **πŸ“ Rule / Config** β€” A short instructions file that shapes how a coding agent behaves in your project β€” like a house-style guide the AI reads before it works (e.g. .cursorrules, AGENTS.md, CLAUDE.md). - **🧷 Extension** β€” An add-on you install into an app or code editor (like a VS Code or Cursor extension) to add AI features to that app. - **πŸš€ App** β€” A standalone AI product you use on its own β€” its own website or app (like an AI video maker or a data-analysis tool). You don't install it into another AI; you just sign up and use it. ### πŸ’¬ vs ⌨️ β€” "chat" tools vs "code" tools Some of these work in **chat apps**, some only in **coding agents**, and some in both: - **πŸ’¬ Chat** β€” Works in AI chat apps like Claude Desktop, ChatGPT, or Gemini β€” the kind where you type a message and it replies. - **⌨️ Code** β€” Works in AI coding agents like Claude Code, Codex, or Cursor β€” tools that can read and edit your files and run commands for you. - **πŸ”€ Both** β€” Works in both chat apps and coding agents. - **🌐 Standalone** β€” Used on its own β€” it has its own website or app, and doesn't run inside Claude, ChatGPT, or Cursor. > This matters! Many of the most popular skills (Superpowers, claude-mem, caveman…) are **⌨️ code-only** β€” they work in Claude Code or Cursor but **not** in a chat app like ChatGPT. Every item below is tagged with where it works. --- ## 🚦 How to use this 1. **Find your role** below (or use the [website](https://rahul-khare997.github.io/ai-toolkit-by-role/) to filter). 2. Start with **⭐ Must-have for everyone**, then your specific role. 3. Check the **Works in** column so you set it up in the right kind of tool. 4. Skip anything marked **🧰 built-in** β€” Claude already ships it. 5. Follow the **name link** for exact, up-to-date install steps. > ⚠️ **Please read:** community-curated, not official docs. The AI space changes fast β€” always confirm install steps and permissions at the linked source before installing, and give connectors the *least* access they need (read-only when possible). Content last checked: 2026-07-13. --- ## πŸ“š Tools by role ### ⭐ Must-have for everyone Start here. These work for almost any job and are the best first things to set up. | Tool | Type | Works in | Cost | Stars | Use it in | What it does & why | |:-----|:-----|:--------:|:----:|:-----:|:----------|:-------------------| | **[Superpowers](https://github.com/obra/superpowers)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 253.4k | Claude Code | **A framework that gives Claude Code a big library of expert 'skills' plus a disciplined workflow (planning, testing, code review).** One install turns Claude Code into a far more capable, methodical developer instead of a one-off helper. | | **[Fetch (read a URL)](https://github.com/modelcontextprotocol/servers/tree/main/src/fetch)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 88.4k | Claude Desktop, Claude Code, Cursor, Codex, Gemini CLI | **Give your AI a link and it fetches the page and reads it.** Perfect for 'summarize this article' or 'pull the key points from this page' without copy-pasting. | | **[Filesystem](https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 88.4k | Claude Desktop, Claude Code, Cursor | **Lets your AI read and write files in folders you choose on your computer.** So a chat app can open your documents and save results. (Coding agents like Claude Code and Cursor can already touch files, so this matters most for chat apps like Claude Desktop.) | | **[Memory](https://github.com/modelcontextprotocol/servers/tree/main/src/memory)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 88.4k | Claude Desktop, Claude Code, Cursor | **Gives your AI a long-term memory that survives between conversations.** So you don't re-explain your project, preferences, or facts every time. It remembers what matters. | | **[Sequential Thinking](https://github.com/modelcontextprotocol/servers/tree/main/src/sequentialthinking)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 88.4k | Claude Desktop, Claude Code, Cursor | **Helps the AI break a hard problem into clear, ordered steps before answering.** Noticeably better results on tricky, multi-step tasks β€” planning, debugging, reasoning. | | **[claude-mem (persistent memory)](https://github.com/thedotmack/claude-mem)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 87k | Claude Code, Cursor, Codex | **Gives your coding agent a memory across sessions by capturing and compressing what it did.** Stop re-explaining your project every session β€” the agent remembers past work and decisions. | | **[GitHub](https://github.com/github/github-mcp-server)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 31.4k | Claude Desktop, Claude Code, Cursor, Codex, GitHub Copilot | **Lets your AI read and manage your GitHub β€” code, issues, and pull requests.** Say 'summarize these open issues' or 'open a PR with this fix' instead of clicking around GitHub yourself. | | **[Claude Code Plugins](https://docs.claude.com/en/docs/claude-code/plugins)** βœ“ | 🧩 Plugin | ⌨️ Code | 🟒 Free | β€” | Claude Code | **Installable bundles that add a set of skills, slash-commands, and connectors to Claude Code at once.** Instead of wiring things up one by one, install a ready-made bundle (for code review, memory, and more) with a single command. | | **[Custom GPTs](https://help.openai.com/en/articles/8554397-creating-a-gpt)** βœ“ | 🧩 Plugin | πŸ’¬ Chat | πŸ”΅ Freemium | β€” | ChatGPT | **Custom versions of ChatGPT that bundle instructions, knowledge, and actions for a specific job.** Set up a reusable assistant once (e.g. 'my brand-voice writer') and everyone can use it β€” no prompt engineering each time. | | **[Web Search (Brave / Exa)](https://modelcontextprotocol.io/examples)** | πŸ”Œ Connector (MCP) | πŸ”€ Both | πŸ”΅ Freemium | β€” | Claude Desktop, Claude Code, Cursor, Codex | **Lets your AI search the live internet and read current web pages.** Out of the box, most AIs only know what they were trained on. This lets them look things up so answers are current and less likely to be made up. | ### πŸ’» Software Developer Write, review, test, and ship code faster with the right connectors and rules. | Tool | Type | Works in | Cost | Stars | Use it in | What it does & why | |:-----|:-----|:--------:|:----:|:-----:|:----------|:-------------------| | **[Superpowers](https://github.com/obra/superpowers)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 253.4k | Claude Code | **A framework that gives Claude Code a big library of expert 'skills' plus a disciplined workflow (planning, testing, code review).** One install turns Claude Code into a far more capable, methodical developer instead of a one-off helper. | | **[Everything Claude Code (ECC)](https://github.com/affaan-m/ECC)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 229.1k | Claude Code, Codex, Cursor | **A power-user setup for coding agents: bundled skills, memory, security checks, and research-first habits.** A tuned configuration that makes agents faster and more reliable on real projects β€” for people past the basics. | | **[Hermes Agent](https://github.com/NousResearch/hermes-agent)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 214k | Claude Code | **An open agent framework designed to adapt and improve as you work with it.** A widely-starred community framework for building agents that get better over time. | | **[gstack (startup-team setup)](https://github.com/garrytan/gstack)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 121.6k | Claude Code | **A bundle of 23 opinionated tools that act like a startup team β€” CEO, designer, eng manager, and QA roles.** A ready-made 'team in a box' for building products with Claude Code, based on a well-known founder's setup. | | **[shadcn/ui components](https://ui.shadcn.com)** βœ“ | πŸ”Œ Connector (MCP) | ⌨️ Code | 🟒 Free | ⭐ 119k | Claude Code, Cursor, Codex | **Lets your coding agent browse and add polished, ready-made UI components.** Build good-looking interfaces fast using proven components instead of styling everything from scratch. | | **[cc-switch (agent manager)](https://github.com/farion1231/cc-switch)** | 🧷 Extension | ⌨️ Code | 🟒 Free | ⭐ 116.6k | Claude Code, Codex | **A desktop app to manage and switch between Claude Code, Codex, and OpenCode setups in one place.** Handy if you use more than one coding agent and want a single control panel. | | **[Caveman (token saver)](https://github.com/JuliusBrussee/caveman)** | 🧠 Skill | ⌨️ Code | 🟒 Free | ⭐ 88.8k | Claude Code | **Makes the AI reply in an ultra-compressed 'caveman' style to save tokens (and money).** Cuts token use by roughly 65% on long sessions with no loss of technical accuracy. | | **[Filesystem](https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 88.4k | Claude Desktop, Claude Code, Cursor | **Lets your AI read and write files in folders you choose on your computer.** So a chat app can open your documents and save results. (Coding agents like Claude Code and Cursor can already touch files, so this matters most for chat apps like Claude Desktop.) | | **[SQLite](https://github.com/modelcontextprotocol/servers/tree/main/src/sqlite)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 88.4k | Claude Desktop, Claude Code, Cursor | **Lets your AI read and query a local SQLite database file.** A simple, no-server way to let the AI explore structured data on your machine. | | **[PostgreSQL / SQL Database](https://github.com/modelcontextprotocol/servers/tree/main/src/postgres)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 88.4k | Claude Desktop, Claude Code, Cursor, Codex | **Connects your AI to a SQL database so it can run queries and read results.** Ask questions in plain English ('how many orders last month?') and let the AI write and run the SQL. | | **[claude-mem (persistent memory)](https://github.com/thedotmack/claude-mem)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 87k | Claude Code, Cursor, Codex | **Gives your coding agent a memory across sessions by capturing and compressing what it did.** Stop re-explaining your project every session β€” the agent remembers past work and decisions. | | **[Graphify (code/docs knowledge graph)](https://github.com/safishamsi/graphify)** | 🧠 Skill | ⌨️ Code | 🟒 Free | ⭐ 83.7k | Claude Code, Codex, Cursor | **Turns a folder of code, SQL schemas, or docs into a queryable knowledge graph the AI can explore.** Ask questions across a whole codebase or dataset at once, instead of file by file. | | **[Context7 (up-to-date docs)](https://github.com/upstash/context7)** βœ“ | πŸ”Œ Connector (MCP) | ⌨️ Code | 🟒 Free | ⭐ 59k | Claude Code, Cursor, Codex, Windsurf | **Feeds your coding agent the latest official documentation for libraries and frameworks.** Stops the AI from using outdated or made-up code β€” it pulls the real, current docs for the exact version you use. | | **[Sub-agents collection](https://github.com/wshobson/agents)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 37.9k | Claude Code | **A marketplace of specialized sub-agents (code reviewer, debugger, architect) for Claude Code.** Drop in expert helpers for specific jobs instead of relying on one generalist agent. | | **[Playwright (browser control)](https://github.com/microsoft/playwright-mcp)** βœ“ | πŸ”Œ Connector (MCP) | ⌨️ Code | 🟒 Free | ⭐ 35k | Claude Code, Cursor, Codex | **Lets your AI open a real browser to click, type, and take screenshots.** Great for testing your website, filling forms, or grabbing screenshots β€” the AI actually drives the browser. | | **[GitHub](https://github.com/github/github-mcp-server)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 31.4k | Claude Desktop, Claude Code, Cursor, Codex, GitHub Copilot | **Lets your AI read and manage your GitHub β€” code, issues, and pull requests.** Say 'summarize these open issues' or 'open a PR with this fix' instead of clicking around GitHub yourself. | | **[Claude Code Templates](https://github.com/davila7/claude-code-templates)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 29.3k | Claude Code | **A CLI that sets up and monitors Claude Code with ready-made configs and dashboards.** Skip manual setup β€” get a solid Claude Code configuration and health view in one command. | | **[Task Master](https://github.com/eyaltoledano/claude-task-master)** | 🧩 Plugin | ⌨️ Code | 🟒 Free | ⭐ 27.8k | Claude Code, Cursor, Codex | **An AI task-management system that breaks a project into tasks your agent works through one by one.** Keeps big builds organized and on-track so the agent ships step by step instead of wandering. | | **[Design-Review Workflows](https://github.com/OneRedOak/claude-code-workflows)** | πŸ“ Rule / Config | ⌨️ Code | 🟒 Free | ⭐ 3.9k | Claude Code | **Battle-tested Claude Code workflows and configs, including strong automated design-review passes.** Copy proven setups (like a rigorous UI review) instead of inventing your own from scratch. | | **[Stripe (payments)](https://github.com/stripe/agent-toolkit)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 1.7k | Claude Desktop, Claude Code, Cursor | **Connects your AI to Stripe to look up payments, customers, and revenue.** Ask about your sales, refunds, or a specific customer in plain English β€” or have a coding agent wire up Stripe for you. | | **[Sentry (error tracking)](https://github.com/getsentry/sentry-mcp)** βœ“ | πŸ”Œ Connector (MCP) | πŸ”€ Both | 🟒 Free | ⭐ 769 | Claude Desktop, Claude Code, Cursor | **Pulls your app's real error reports into the AI.** Ask 'what's breaking in production and why?' and get the AI to read the actual stack traces and suggest fixes. | | **[Semgrep (security scan)](https://github.com/semgrep/mcp)** βœ“ | πŸ”Œ Connector (MCP) | ⌨️ Code | 🟒 Free | ⭐ 680 | Claude Code, Cursor | **Scans your code for security bugs and lets the AI explain and fix them.** Catches vulnerabilities as you build, in plain language, instead of after something goes wrong. | | **[Agent Instructions (CLAUDE.md / AGENTS.md)](https://agents.md)** βœ“ | πŸ“ Rule / Config | ⌨️ Code | 🟒 Free | β€” | Claude Code, Codex, Cursor, Gemini CLI | **A plain-text file your coding agent reads first, describing your project and how to work in it.** One place to set context and rules β€” the agent stops guessing and works the way your team does. | | **[Bolt](https://bolt.new)** βœ“ | πŸš€ App | 🌐 Standalone | πŸ”΅ Freemium | β€” | Its own app | **Prompt-to-app builder that runs entirely in the browser.** Spin up and preview a working web app in seconds. | | **[Claude Code Plugins](https://docs.claude.com/en/docs/claude-code/plugins)** βœ“ | 🧩 Plugin | ⌨️ Code | 🟒 Free | β€” | Clau … ## Source & license This open-source MCP server is cataloged on AgentStack and links to its original source β€” we do not rehost the code. - **Author:** [Rahul-Khare997](https://github.com/Rahul-Khare997) - **Source:** [Rahul-Khare997/ai-toolkit-by-role](https://github.com/Rahul-Khare997/ai-toolkit-by-role) - **License:** MIT Install and usage instructions live in the source repository linked above. ## Pricing - **Free** β€” Free ## Security capabilities Automated source analysis of v0.1.0 β€” what this tool can access: - **Network access:** yes - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source β€” more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** β€” security scan: passed β€” Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/mcp-rahul-khare997-ai-toolkit-by-role - Seller: https://agentstack.voostack.com/s/rahul-khare997 - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack β€” the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.