How to set up Claude Code: add-ons worth installing

Before you install anything

Add-ons come in two shapes. MCP servers connect Claude Code to an outside tool and are added with claude mcp add in your terminal. Plugins bundle skills, commands, agents or MCP servers and are installed inside Claude Code with /plugin install <name>@<marketplace>. Claude Code adds Anthropic's official marketplace, claude-plugins-official, for you the first time you start an interactive terminal session. MCP servers you add are stored at local scope (this project, only you) by default; add --scope project to share one with your team through a .mcp.json file, or --scope user to load it in all your projects. Run claude mcp list to check each server's status. Anthropic's plugin directory asks you to make sure you trust a plugin before installing it, because Anthropic doesn't control what MCP servers, files or other software a plugin includes.

Source: code.claude.com/docs/en/mcp, code.claude.com/docs/en/discover-plugins, github.com/anthropics/claude-plugins-official

1. Context7: current library docs

By Upstash · github.com/upstash/context7

Context7 pulls up-to-date, version-specific documentation and code examples for the libraries you use into Claude Code, so it writes against the current API instead of what it remembers from training. It works either as a CLI plus a skill, or as an MCP server. Add "use context7" to a prompt, or put a rule in CLAUDE.md telling Claude to use it whenever it needs library docs.

npx ctx7 setup --claude

claude mcp add --scope user --header "Authorization: Bearer YOUR_API_KEY" --transport http context7 https://mcp.context7.com/mcp

Context7 recommends a free API key from context7.com/dashboard for higher rate limits. The guided setup signs you in and creates the key for you.

Skip it if: You mostly work with one stable, well-known stack and rarely hit outdated or made-up APIs.

Source: github.com/upstash/context7, context7.com/docs/resources/all-clients

2. Playwright MCP: a real browser

By Microsoft · github.com/microsoft/playwright-mcpPlaywright on StackShip

Playwright MCP lets Claude Code open and drive a browser: navigate, click, fill forms and read pages. It works from Playwright's accessibility tree rather than screenshots, so it doesn't need a vision model. Useful for checking that the feature Claude just built actually works in a browser. It requires Node.js 18 or newer.

claude mcp add playwright npx @playwright/mcp@latest

Skip it if: You want the lightest option for a coding agent. The Playwright MCP README itself points coding agents to Playwright CLI with skills, because CLI calls use fewer tokens than loading MCP tool schemas and accessibility trees.

Source: github.com/microsoft/playwright-mcp, github.com/microsoft/playwright-cli

3. GitHub MCP server: issues and PRs

By GitHub · github.com/github/github-mcp-server

GitHub's official MCP server gives Claude Code access to your repositories, issues and pull requests, and search across GitHub. The remote server is hosted by GitHub; you authenticate with a GitHub personal access token. There's also a local version that runs in Docker and can sign you in through the browser instead.

claude mcp add-json github '{"type":"http","url":"https://api.githubcopilot.com/mcp","headers":{"Authorization":"Bearer YOUR_GITHUB_PAT"}}'

claude mcp list

Keep the token out of your repo. GitHub's guide suggests putting it in .env and adding .env and .mcp.json to .gitignore.

Skip it if: Claude Code already does what you need through the GitHub CLI (gh), or you don't want a personal access token stored in your MCP config.

Source: github.com/github/github-mcp-server, github.com/github/github-mcp-server/blob/main/docs/installation-guides/install-claude.md

4. Sentry MCP: production errors

By Sentry · github.com/getsentry/sentry-mcpSentry on StackShip

Sentry's MCP server lets Claude Code look up your Sentry errors, issues, traces and performance data while it debugs. Sentry ships it as a Claude Code plugin that adds a sentry-mcp subagent, which Claude hands Sentry questions to automatically.

claude plugin marketplace add getsentry/sentry-mcp

claude plugin install sentry-mcp@sentry-mcp

Skip it if: You don't use Sentry for error monitoring yet.

Source: github.com/getsentry/sentry-mcp

5. code-review: a second look at every PR

By Anthropic · github.com/anthropics/claude-plugins-official/tree/main/plugins/code-review

An official Anthropic plugin that adds a /code-review command. It launches four review agents in parallel on a pull request: two check it against your CLAUDE.md rules, one looks for obvious bugs in the changes, and one reads git history for context. Each issue gets a confidence score, and only issues scored 80 or higher are reported, to cut false positives. It needs the GitHub CLI (gh) installed and signed in.

/plugin install code-review@claude-plugins-official

/code-review

Skip it if: You don't work through pull requests, or your project is too small to need a second review.

Source: github.com/anthropics/claude-plugins-official/tree/main/plugins/code-review, code.claude.com/docs/en/discover-plugins

6. Superpowers: a structured build workflow

By obra (community) · github.com/obra/superpowers

Superpowers is a set of skills that changes how Claude Code approaches a build. Instead of jumping straight into code, it asks what you're trying to do, writes a spec you sign off in short chunks, turns that into an implementation plan, then builds with red/green test-driven development and subagents that review each other's work. The skills trigger automatically. It's a community project, listed in Anthropic's official plugin marketplace.

/plugin install superpowers@claude-plugins-official

Skip it if: Most of your work is quick fixes and small edits, where a spec and plan step slows you down more than it helps.

Source: github.com/obra/superpowers, github.com/anthropics/claude-plugins-official

7. Anthropic's skills: documents and examples

By Anthropic · github.com/anthropics/skills

Anthropic's skills repo is a plugin marketplace with two sets: document-skills (working with Word, PDF, PowerPoint and Excel files) and example-skills (creative, technical and enterprise examples). Skills are folders of instructions and scripts that Claude loads when a task needs them. Mention a skill in your prompt to use it, for example "use the PDF skill to extract the form fields from this file". It's also a good starting point for writing your own skills.

/plugin marketplace add anthropics/skills

/plugin install document-skills@anthropic-agent-skills

The other set installs with /plugin install example-skills@anthropic-agent-skills. Anthropic says these skills are for demonstration and education, so test them before relying on them. The document skills are source-available, not open source.

Skip it if: You never handle Office or PDF files in your projects and don't plan to write your own skills.

Source: github.com/anthropics/skills

8. Claude Code GitHub Action: @claude in PRs and issues

By Anthropic · github.com/anthropics/claude-code-actionClaude Code on StackShip

Anthropic's GitHub Action runs Claude Code inside your GitHub workflows. Mention @claude in a pull request or issue and it can answer questions, review the changes or implement a fix. It runs on your own GitHub runner.

/install-github-app

You need to be a repository admin to install the GitHub app and add secrets. The quick setup only works if you use the Anthropic API directly; Amazon Bedrock, Google Vertex AI and Microsoft Foundry have separate setup docs.

Skip it if: You only run Claude Code locally and don't want it acting on your repo from GitHub.

Source: github.com/anthropics/claude-code-action