The full 8+8+8 list of Claude plug-ins, skills, and MCP servers — each with its repo URL and exact install command, so you can build a fully loaded Claude in one sitting.
This guide is reviewed for clarity, service accuracy, and AI-search readability. The next quarterly content review is tracked internally before unsupported metrics or client proof are added.
24 Things to Install in Claude: The Full 8 + 8 + 8 List
Most people use Claude like a search box. A few of us run it like a full team.
This is the complete install list: 8 plug-ins, 8 skills, and 8 MCP servers. Plug-ins drop in whole specialist teams. Skills turn a 200-word prompt into an 8-character command. MCP servers wire Claude into the apps your business already runs on, so it does the work instead of describing it.
You can set up the whole stack in one sitting. Every item below has what it does, when to reach for it, the repo or marketplace URL, and the exact command to install it. Copy, paste, done.
Start Here: 15-Minute Setup
Before you install anything, get the base right.
Prerequisites
- Claude Code (the CLI) — the terminal version where plug-ins, skills, and MCP all work. Install or update it:
bashnpm install -g @anthropic-ai/claude-code claude --version # confirm it installed
- Node.js 18+ and npm — most MCP servers run through
npx. Check withnode --version. - Python runner
uv(optional) — a few MCP servers run throughuvx. Install it if you want them:
bashcurl -LsSf https://astral.sh/uv/install.sh | sh
- Claude Desktop (optional) — the app version. It reads MCP servers from a file called
claude_desktop_config.json. Use Desktop if you prefer a chat window over a terminal. Everything in Section 3 works in both.
Plug-ins vs Skills vs MCP — know the difference
These three words get mixed up constantly. Here is the clean version:
- Plug-in = a bundle. One install can ship commands, sub-agents, skills, and even MCP servers together. Think of it as hiring a whole team in one command.
- Skill = a single folder with a
SKILL.mdfile that teaches Claude one workflow. It loads automatically when it is relevant, or fires on a slash command. No API keys, no network — just instructions (and sometimes a helper script). - MCP server = a live connector to an outside app. It gives Claude real tools to read and write that app's data — your GitHub, your Stripe, your database. This is the one that needs an API key or a login.
Install order
Do it in this order and nothing trips over itself:
- Plug-ins — fastest, no keys needed.
- Skills — drop-in folders, no keys needed.
- MCP servers — last, because these need API keys and logins.
At Hypd we run this exact stack daily. Let's build it.
Section 1 — 8 Plug-ins (Whole Teams in One Command)
Plug-ins come from a marketplace — a repo Claude reads to find installable bundles. Anthropic runs the official one. You add the marketplace once, then install any plug-in from it.
bash# Add the official directory one time /plugin marketplace add anthropics/claude-plugins-official # Then install any plug-in from it /plugin install <name>@claude-plugins-official # Or browse the live catalog visually /plugin
Plug-in names can change as the directory grows. The marketplace command above never changes — if an install name has moved, type /plugin and the browser shows the current slug. Today's directory skews toward engineering teams (that's what Claude Code is built for); for finance, legal, and marketing "teams," you'll lean on the skills and MCP connectors in Sections 2 and 3.
Here are 8 worth installing first.
1. Code Review — your senior reviewer. Reads your diff and flags real bugs, edge cases, and risky changes before you ship.
bash/plugin install code-review@claude-plugins-official
2. Security Review — your appsec auditor. Scans changes for injection holes, leaked secrets, and broken auth. Source: github.com/anthropics/claude-code-security-review.
bash/plugin install security-review@claude-plugins-official
3. Frontend Design — your product designer. Builds UI that doesn't look like a template, with real typography and layout choices.
bash/plugin install frontend-design@claude-plugins-official
4. PR Review — your pull-request manager. Summarizes, labels, and reviews PRs so your repo stays clean solo.
bash/plugin install pr-review@claude-plugins-official
5. Docs Writer — your technical writer. Turns code into READMEs, changelogs, and API docs you'd actually publish.
bash/plugin install docs@claude-plugins-official
6. Test Generator — your QA engineer. Writes and runs tests for the code you just changed.
bash/plugin install testing@claude-plugins-official
7. Debugger — your root-cause investigator. Traces a failing test or stack trace back to the real cause instead of guessing.
bash/plugin install debugger@claude-plugins-official
8. Commit Assistant — your git hand. Writes clean conventional commits and manages branches so your history stays readable.
bash/plugin install commit@claude-plugins-official
Section 2 — 8 Skills (One Command Fires a Whole Workflow)
A skill is a folder with a SKILL.md file. Personal skills live in ~/.claude/skills/. Claude loads a skill on its own when the work matches it, or you fire it directly. No keys, no setup beyond dropping the folder in place.
Anthropic publishes a pack of them in one repo, so most of this section installs from a single clone.
bash# Clone the official skills repo once git clone https://github.com/anthropics/skills.git ~/anthropic-skills # Copy any skill folder you want into your personal skills dir cp -r ~/anthropic-skills/skill-creator ~/.claude/skills/skill-creator
After copying, restart Claude Code so it picks up the new skill.
1. Skill Creator — the meta-skill. Describe a repetitive job out loud and it scaffolds a brand-new installed skill for it. Fires on /skill-creator or "make me a skill that…". Path: ~/.claude/skills/skill-creator (from github.com/anthropics/skills).
2. MCP Builder — scaffolds a working MCP server from a plain-English spec, so you can connect a tool that doesn't have a server yet. Fires on "build an MCP server for…". Path: ~/.claude/skills/mcp-builder (from github.com/anthropics/skills).
3. PDF — fills, merges, splits, and extracts text from PDFs. Turn "fill out this form and pull the totals" into one step. Fires when you hand Claude a PDF. Path: ~/.claude/skills/pdf (from github.com/anthropics/skills).
4. XLSX — builds and edits real spreadsheets with working formulas, not flat text. Great for pricing models and reports. Fires on "build a spreadsheet that…". Path: ~/.claude/skills/xlsx (from github.com/anthropics/skills).
5. PPTX — generates branded slide decks from an outline. Fires on "make a deck about…". Path: ~/.claude/skills/pptx (from github.com/anthropics/skills).
6. DOCX — drafts and edits Word documents — contracts, proposals, one-pagers — keeping styles intact. Fires on "write me a doc for…". Path: ~/.claude/skills/docx (from github.com/anthropics/skills).
7. Superpowers — a whole library of workflow skills (planning, test-driven development, debugging, brainstorming). It ships as its own plug-in marketplace:
bash/plugin marketplace add obra/superpowers /plugin install superpowers@superpowers
Repo: github.com/obra/superpowers.
8. Firecrawl — gives Claude reliable web search, scraping, and browser automation as a skill, so it can research and pull live data on its own. Drop in the skill and set one key:
bashgit clone https://github.com/firecrawl/firecrawl.git ~/firecrawl cp -r ~/firecrawl/agent-onboarding ~/.claude/skills/firecrawl export FIRECRAWL_API_KEY="fc-your-key" # from firecrawl.dev
Repo: github.com/firecrawl/firecrawl.
Section 3 — 8 MCP Servers (Claude's Live Operations Layer)
This is the real shift. An MCP server turns Claude from a chat window into something that reads and writes your actual tools.
In Claude Code you add a server with claude mcp add. Two flavors:
- Local servers run on your machine through
npx:claude mcp add <name> -- npx -y <package> - Remote servers are hosted by the company; you connect over HTTP or SSE and log in:
claude mcp add --transport http <name> <url>
Check what's connected anytime with claude mcp list. The official server directory is github.com/modelcontextprotocol/servers, and the public registry lives at modelcontextprotocol.io.
1. Filesystem — reads and writes files in a folder you choose. The foundation: Claude can actually open and edit your project. Package: @modelcontextprotocol/server-filesystem (github.com/modelcontextprotocol/servers). No key — just point it at a folder.
bashclaude mcp add filesystem -- npx -y @modelcontextprotocol/server-filesystem ~/projects
2. GitHub — reads and writes your repos, issues, and pull requests. Connect to the official hosted server and log in once. Repo: github.com/github/github-mcp-server.
bashclaude mcp add --transport http github https://api.githubcopilot.com/mcp/
3. Notion — reads and writes your Notion workspace: docs, databases, tasks. Official hosted server; you authorize on connect. Repo: github.com/makenotion/notion-mcp-server.
bashclaude mcp add --transport http notion https://mcp.notion.com/mcp
4. Stripe — reads payments, customers, and invoices, and can create them. Your revenue layer, live. Official hosted server. Repo: github.com/stripe/agent-toolkit.
bashclaude mcp add --transport http stripe https://mcp.stripe.com
5. Supabase — reads and writes your Postgres database, runs queries, and inspects schema. Needs a personal access token (sbp_…) from your Supabase dashboard. Repo: github.com/supabase-community/supabase-mcp.
bashclaude mcp add supabase -- npx -y @supabase/mcp-server-supabase@latest --access-token sbp_your_token
6. Linear — reads and writes issues, projects, and cycles. Your project tracker becomes something Claude updates for you. Official hosted server; authorize on connect.
bashclaude mcp add --transport sse linear https://mcp.linear.app/sse
7. Playwright — drives a real browser: clicks, fills forms, screenshots, and tests live pages. Your hands on the web. Package: @playwright/mcp (github.com/microsoft/playwright-mcp). No key.
bashclaude mcp add playwright -- npx -y @playwright/mcp@latest
8. Sentry — reads your error and performance data so Claude can debug production from the actual stack traces. Official hosted server; authorize on connect. Repo: github.com/getsentry/sentry-mcp.
bashclaude mcp add --transport http sentry https://mcp.sentry.dev/mcp
Copy-Paste Config (Claude Desktop)
If you use Claude Desktop instead of the CLI, you wire MCP servers in a JSON file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%Claudeclaude_desktop_config.json
Paste this, swap in your keys, and restart the app. Local servers go straight in; hosted servers connect through the mcp-remote helper.
json{ "mcpServers": { "filesystem": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/projects"] }, "supabase": { "command": "npx", "args": ["-y", "@supabase/mcp-server-supabase@latest"], "env": { "SUPABASE_ACCESS_TOKEN": "sbp_your_token" } }, "playwright": { "command": "npx", "args": ["-y", "@playwright/mcp@latest"] }, "github": { "command": "npx", "args": ["-y", "mcp-remote", "https://api.githubcopilot.com/mcp/"] }, "notion": { "command": "npx", "args": ["-y", "mcp-remote", "https://mcp.notion.com/mcp"] }, "stripe": { "command": "npx", "args": ["-y", "mcp-remote", "https://mcp.stripe.com"] }, "linear": { "command": "npx", "args": ["-y", "mcp-remote", "https://mcp.linear.app/sse"] }, "sentry": { "command": "npx", "args": ["-y", "mcp-remote", "https://mcp.sentry.dev/mcp"] } } }
Keys each server needs
- Filesystem — none. Just the folder path in the args.
- Playwright — none.
- GitHub — log in on first connect (OAuth). For a local-only setup, use
GITHUB_PERSONAL_ACCESS_TOKENfrom your GitHub developer settings. - Notion / Linear / Sentry — none to paste. A browser window opens to authorize on first connect.
- Stripe — log in on connect, or set
STRIPE_SECRET_KEY(sk_…) from your Stripe dashboard for local use. - Supabase —
SUPABASE_ACCESS_TOKEN(sbp_…) from Supabase → Account → Access Tokens.
Keep every key in this file or your shell environment — never paste secrets into a chat.
All 24 at a Glance
Scan this, pick what fits your week, and install just those.
| # | Name | Type | What it does | Install |
|---|---|---|---|---|
| 1 | Code Review | Plug-in | Senior reviewer for your diffs | /plugin install code-review@claude-plugins-official |
| 2 | Security Review | Plug-in | Appsec audit on changes | /plugin install security-review@claude-plugins-official |
| 3 | Frontend Design | Plug-in | Non-templated UI design | /plugin install frontend-design@claude-plugins-official |
| 4 | PR Review | Plug-in | Summarizes and reviews PRs | /plugin install pr-review@claude-plugins-official |
| 5 | Docs Writer | Plug-in | Code → READMEs and changelogs | /plugin install docs@claude-plugins-official |
| 6 | Test Generator | Plug-in | Writes and runs tests | /plugin install testing@claude-plugins-official |
| 7 | Debugger | Plug-in | Traces failures to root cause | /plugin install debugger@claude-plugins-official |
| 8 | Commit Assistant | Plug-in | Clean commits and branches | /plugin install commit@claude-plugins-official |
| 9 | Skill Creator | Skill | Builds new skills for you | cp -r ~/anthropic-skills/skill-creator ~/.claude/skills/ |
| 10 | MCP Builder | Skill | Scaffolds MCP servers | cp -r ~/anthropic-skills/mcp-builder ~/.claude/skills/ |
| 11 | Skill | Fill, merge, extract PDFs | cp -r ~/anthropic-skills/pdf ~/.claude/skills/ | |
| 12 | XLSX | Skill | Spreadsheets with formulas | cp -r ~/anthropic-skills/xlsx ~/.claude/skills/ |
| 13 | PPTX | Skill | Branded slide decks | cp -r ~/anthropic-skills/pptx ~/.claude/skills/ |
| 14 | DOCX | Skill | Contracts and proposals | cp -r ~/anthropic-skills/docx ~/.claude/skills/ |
| 15 | Superpowers | Skill | Planning/TDD/debug library | /plugin install superpowers@superpowers |
| 16 | Firecrawl | Skill | Web search + scrape | clone firecrawl/firecrawl → skills dir |
| 17 | Filesystem | MCP | Read/write local files | claude mcp add filesystem -- npx -y @modelcontextprotocol/server-filesystem ~/projects |
| 18 | GitHub | MCP | Repos, issues, PRs | claude mcp add --transport http github https://api.githubcopilot.com/mcp/ |
| 19 | Notion | MCP | Docs and databases | claude mcp add --transport http notion https://mcp.notion.com/mcp |
| 20 | Stripe | MCP | Payments and invoices | claude mcp add --transport http stripe https://mcp.stripe.com |
| 21 | Supabase | MCP | Postgres database | claude mcp add supabase -- npx -y @supabase/mcp-server-supabase@latest --access-token sbp_… |
| 22 | Linear | MCP | Issues and projects | claude mcp add --transport sse linear https://mcp.linear.app/sse |
| 23 | Playwright | MCP | Live browser control | claude mcp add playwright -- npx -y @playwright/mcp@latest |
| 24 | Sentry | MCP | Production error data | claude mcp add --transport http sentry https://mcp.sentry.dev/mcp |
FAQ
Do I need all 24 on day one? No. Install the four foundations first — the Filesystem MCP, the Code Review plug-in, the Skill Creator skill, and one MCP for an app you live in (GitHub, Notion, or Stripe). Add the rest as a real task asks for them. The list is a menu, not a checklist.
A plug-in install name didn't work — what happened?
The directory updates over time and a slug may have been renamed. Run /plugin to open the live catalog and copy the current name. The /plugin marketplace add anthropics/claude-plugins-official step never changes, so you only ever fix the install line, not the setup.
Is it safe to connect MCP servers to my real Stripe and database? Yes, with normal care. Use the official hosted servers or the official packages listed here, keep API keys in your config file or environment (never in chat), and start any write-capable server — like Supabase or Stripe — with a limited-scope or read-only key until you trust the workflow. MCP servers only do what their key allows, so scope the key and you've scoped Claude.
This is the stack we run at Hypd — specialist teams, one-line workflows, and a live operations layer. Set up a few before Monday and let Claude do the work instead of describing it.