Best Claude Code GitHub Repos (2026): 15 to Install

The best Claude Code GitHub repos right now: Ay-Skills for plug-and-play production skills, Superpowers for TDD-enforced development, Everything Claude Code for a full agent harness, and claude-code-action for CI/CD. Which one you install first depends on whether you need skills, a framework, or infrastructure.

Last updated: November 10, 2026. Every star count below was read from the GitHub API on 2026-11-10.

The scale explains the problem. Claude Code has gone from a niche terminal tool to the backbone of how serious developers ship code in 2026, and the repos that get things done are buried under noise.

This guide covers the 15 best GitHub repositories for Claude Code: plug-and-play skill packs, TDD frameworks, Anthropic's official skills and plugin repos, MCP servers, complete agent harnesses, a usage tracker and two repos released since mid-October. Star counts were re-read through the GitHub API on 2026-11-10; licenses and releases were re-checked the same day where the text gives a date. We break down what each repo does, who it's for, and whether it's worth adding to your workflow.

What changed since Claude Fable 5.2 launched

Anthropic released Claude Fable 5.2 in Claude Code 2.1.321 on 2026-11-09 and made it the default Fable model (claude-fable-5-2, 1M context, $8 input and $40 output per million tokens, per the Claude Code changelog). The repos on this list do not depend on the model, but a few things moved around that release:

• The official Claude Code repo kept growing. anthropics/claude-code had 155,154 stars on 2026-11-10, up from 152,683 on 2026-10-21. If you searched for "the Claude repo", that is the one: issue tracker, changelog and release notes.
• Anthropic's plugins now live in one repo. On 2026-10-26 Anthropic deleted the stale plugin copies in anthropics/claude-code and pointed its demo marketplace at anthropics/claude-plugins-official. That repo is new to this list at #13.
• Two new repos earned a place. Theo Browne's tstack (created 2026-10-24) and Floorplan (created 2026-10-27) each passed 3,400 stars in about two weeks. They are #14 and #15.
• Status changes. claude-code-mcp is still archived. awesome-claude-code-toolkit has had no push since 2026-05-12. Superpowers shipped v7.0.2 on 2026-11-03 and ccusage shipped v20.0.48 on 2026-11-02.

Best Claude Code GitHub repos: a brief overview

If you want production-ready skills you can drop in today:

• Start with Ay-Skills for 10 skills covering SEO, browser automation, video, and diagrams
• Add Superpowers when you want strict TDD enforcement across your entire dev workflow
• Read anthropics/skills for Anthropic's own reference skills and the Agent Skills spec

If you want a complete agent framework:

• Go with Everything Claude Code for the most complete harness (71 agents, security scanning, memory)
• Choose awesome-claude-code-toolkit for a curated toolkit with 400K+ skills via SkillKit (no updates since May 2026)
• Add tstack for Theo Browne's PR-shipping skills: full-send, council, babysit-pr and six more

The agent frameworks space moves fast. For an autonomous agent framework that crossed 250K GitHub stars in under 60 days, see our overview of OpenClaw.

If you need CI/CD and infrastructure integration:

• Start with claude-code-action for GitHub Actions automation on PRs and issues
• Add github-mcp-server to give Claude Code native access to repos, PRs, and CI workflows
• Run ccusage to see daily and per-session token usage and estimated cost
• Browse claude-plugins-official, Anthropic's marketplace repo, to read any plugin's source before you install it
• Install Floorplan to get a clickable map of how your app fits together

If you want to learn and explore:

• Use awesome-claude-code as a large directory of plugins and skills
(For a curated roundup instead of a raw directory, see our best Claude Code plugins guide.)
• Study claude-code-ultimate-guide for its guide, runnable examples and templates (430K+ lines per its repo description)

GitHub stars for all 15 repos, in list order

Ay-Skills

94

Superpowers

315,400

Everything Claude Code

300,465

anthropics/skills

191,742

claude-code-action

9,872

github-mcp-server

34,361

awesome-claude-code-toolkit

2,829

awesome-claude-code

57,668

claude-code-mcp

1,315

claude-context

12,683

claude-code-ultimate-guide

6,400

ccusage

19,389

claude-plugins-official

46,697

tstack

3,422

Floorplan

3,404

1. Ay-Skills, best plug-and-play skill pack for Claude Code

Ay-Skills is a public collection of Claude Code skills built by AY Automate. Unlike most skill repos that offer a handful of narrow utilities, Ay-Skills ships 10 production-grade skills that cover a wide range of real workflows: SEO audits, headless browser automation, Remotion video creation, Excalidraw diagramming, MCP client connections, and more.

Each skill is a self-contained folder you drop into .claude/skills/. No configuration, no dependencies to manage. Claude Code auto-loads it and gains the capability immediately. The repo stands out because every skill was built for production use at an agency, not as a demo or proof of concept.

Why it's on the list

• A public collection of 10 Claude Code skills, each in its own skills/ folder: artifacts-builder, mcp-client, remotion, skill-creator, notebooklm, ui-ux-pro-max, agent-browser, claude-seo, superpowers and excalidraw-diagram.
• Fits Claude Code users at any level who want ready-made skills for SEO audits, browser automation, video, diagrams and MCP. Install by copying a skill folder into .claude/skills/.
• Free and open source under the MIT license.
• Proof: 94 GitHub stars, read from the GitHub API for walidboulanouar/Ay-Skills on 2026-11-10.
• Built by AY Automate, whose AI-native engineers embed in client teams from $3K a month (how we work).

Key features

• 10 production-ready skills: SEO audits, browser automation, Remotion video, Excalidraw diagrams, MCP client, UI/UX design system, NotebookLM integration, artifact builder, skill creator, and Superpowers TDD
• True plug-and-play: copy a folder into .claude/skills/ and it works
• Each skill includes supporting scripts, reference files, and detailed trigger conditions
• Skills cover both technical (code, MCP, browser) and creative (video, diagrams, UI) workflows
• Used by AY Automate on its own client builds (our repo)

Best for

• Developers who want a curated set of high-quality skills rather than sifting through thousands of mediocre ones
• Teams running AI agent development workflows who need browser automation, SEO, and diagramming in their Claude Code setup
• Anyone who wants to see how well-structured Claude Code skills should be built

Pricing

• Free and open source under the MIT license, added to the repo on 2026-09-26
• No paid tiers or premium features

Pros

• Used by AY Automate on its own client builds (our repo)
• Covers an unusually wide range of use cases in a single repo (SEO, video, diagrams, browser, MCP)
• Clean folder structure makes it trivial to install one skill or all ten
• Includes a skill-creator skill for building your own skills following best practices

2. Superpowers, best TDD-enforced dev framework

Superpowers is a development methodology encoded as a skills framework that turns Claude Code from a reactive code generator into something that behaves like a senior developer who follows test-driven development. With 315,400 GitHub stars (2026-11-10) and official acceptance into the Anthropic marketplace, it is the most-starred repo on this list, ahead of Everything Claude Code (300,465). The latest release is v7.0.2, published 2026-11-03.

The framework enforces a strict 7-step workflow, as the README lists it: brainstorming, git worktree setup, writing plans, subagent-driven or inline plan execution, TDD, code review, and finishing the branch. For the sub-agents worth adding to that workflow, see the 10 best Claude Code sub-agents. The key differentiator is that Superpowers literally deletes code written before tests exist. TDD is not optional.

Why it's on the list

• An agentic skills framework and software development methodology that enforces test-driven development across Claude Code sessions.
• Fits engineering teams and solo developers who want one shared, strict workflow. Expect a learning curve.
• Free and open source under the MIT license.
• Proof: 315,400 GitHub stars, read from the GitHub API for obra/superpowers on 2026-11-10, author obra.

Key features

• 7-phase agentic workflow with mandatory TDD enforcement
• 15 skills listed in the README skills library (checked 2026-10-21), covering testing, debugging, collaboration, and planning
• Skills trigger automatically at each step; the repo no longer has the commands folder that held /brainstorm, /write-plan and /execute-plan (checked 2026-10-21)
• YAGNI and DRY principles baked into every skill
• Socratic brainstorming phase forces requirement clarity before any code is written

Best for

• Engineering teams that want to enforce consistent development practices across all Claude Code usage
• Solo developers who want the discipline of a senior code reviewer built into their workflow
• Organizations adopting Claude Code at scale who need a shared methodology

Pricing

• Free and open source
• Available on the Anthropic marketplace

Pros

• TDD enforcement catches bugs before they ship and cuts debugging time
• The structured workflow prevents the common problem of Claude Code generating code before requirements are clear
• Massive community (315K+ stars on 2026-11-10) and a release as recent as 2026-11-03

Cons

• The strict TDD enforcement can feel heavy for quick prototyping or exploratory coding
• Learning curve: the 7-phase workflow requires understanding before you see the benefit
• Opinionated methodology may conflict with teams that have their own established dev practices

3. Everything Claude Code, best complete agent harness system

Everything Claude Code (ECC) is an MIT-licensed agent harness for Claude Code with 300,465 stars (2026-11-10). Its author won the Anthropic x Forum Ventures hackathon in September 2025, and its AgentShield scanner was built at the Claude Code Hackathon (Cerebral Valley x Anthropic, February 2026), per the README. Where Superpowers focuses on TDD methodology, ECC provides a full operating system: 71 agents, security scanning, memory optimization, continuous learning, and production-ready configurations.

The latest GitHub release is still v2.2.3 (October 1, 2026), though commits landed as recently as 2026-11-09. The README says ECC works best in Claude Code, has native Codex and Qoder plugin paths, and ships capability-limited adapters for Cursor, OpenCode, Gemini and other harnesses. Per the README, the repo includes 71 agents, 311 skills and 97 legacy command shims, plus hooks, rules, memory and AgentShield security scanning.

Why it's on the list

• An agent harness for Claude Code that bundles skills, agents, memory, security scanning and research-first development workflows.
• Fits experienced Claude Code users who want a full operating setup rather than a single skill.
• Free and open source under the MIT license.
• Proof: 300,465 GitHub stars, read from the GitHub API for affaan-m/ECC on 2026-11-10.

Key features

• 71 agents, 311 skills and 97 legacy command shims (README, 2026-10-21)
• AgentShield config scanner: 102 static analysis rules, 1,282 tests, 98% coverage (per the README)
• Cross-platform: works with Claude Code, Codex, OpenCode, and Cursor
• Harness commands: /harness-audit, /loop-start, /quality-gate, /model-route

Best for

• Enterprise teams deploying Claude Code across multiple projects who need security scanning and quality gates
• Developers who want a single harness that works across multiple AI coding tools beyond Claude Code
• Teams that need production-grade agent orchestration with monitoring and metrics

Pricing

• Free and open source
• No paid tiers

Pros

• The most complete agent system available, covering security, testing, memory, and multi-model routing
• Cross-platform support means you don't have to choose between Claude Code and Cursor
• Includes the AgentShield config scanner, which the README lists with 102 static analysis rules, 1,282 tests and 98% coverage

Cons

• The sheer scope can be overwhelming for developers who just want a few skills
• Heavy setup compared to simpler repos like Ay-Skills or Superpowers
• Rapid iteration (frequent breaking changes between versions)

4. anthropics/skills, best official reference for Agent Skills

anthropics/skills is Anthropic's public repository for Agent Skills. It holds example skills, the document skills (docx, pdf, pptx, xlsx) behind Claude's file creation, the Agent Skills specification and a starter template. On 2026-11-10 the repo showed 191,742 stars and a last push on 2026-11-06.

Ay-Skills and Superpowers give you finished workflows. This repo shows you the format those workflows are built on. Each skill is a folder with a SKILL.md file, and the frontmatter needs only a name and a description. We placed it fourth because it is the reference every other skills repo on this list follows. For where to find third-party skills beyond GitHub, see our guide to the Claude skills marketplace.

Why it's on the list

• Anthropic's public repository for Agent Skills: example skills, the docx, pdf, pptx and xlsx document skills, the Agent Skills spec and a template.
• Fits developers writing their own skills and teams that need Office file generation in Claude Code.
• Free. Many skills are Apache 2.0; the four document skills are source-available, per the README.
• Proof: 191,742 GitHub stars, read from the GitHub API for anthropics/skills on 2026-11-10, published by Anthropic.

Key features

• Installs as a Claude Code plugin marketplace with /plugin marketplace add anthropics/skills, then pick document-skills or example-skills
• 19 skill folders as of 2026-10-21, including mcp-builder, skill-creator, webapp-testing, frontend-design and the four document skills
• The Agent Skills specification lives in /spec and a template skill in /template
• According to the README, the same skills are available on paid Claude.ai plans and through the Claude API

Best for

• Developers writing their own skills who want the reference format straight from Anthropic
• Teams that need Word, PDF, PowerPoint or Excel generation inside Claude Code

Pricing

• Free. The README says many skills are Apache 2.0, while the docx, pdf, pptx and xlsx skills are source-available, not open source

Pros

• Maintained by Anthropic, so the examples follow the current skill format
• The document skills are, per the README, actively used in a production AI application, which makes them good study material for complex skills
• One command adds the whole set to Claude Code

Cons

• The README labels the skills as provided for demonstration and educational purposes only, and warns that Claude's built-in behavior may differ
• Mixed licensing: you cannot reuse the document skills as freely as the Apache 2.0 ones

Skip it if

• You want an enforced development workflow or a full agent harness. Superpowers or Everything Claude Code fit that better

5. claude-code-action, best GitHub Actions CI/CD integration

claude-code-action is Anthropic's official GitHub Action that runs Claude Code directly in your CI/CD pipeline. It detects @claude mentions in PR comments, responds to issue assignments, and can implement code changes autonomously. This is the bridge between Claude Code on your laptop and Claude Code in your team's automation pipeline.

The action runs the full Claude Code runtime in a GitHub Actions runner, reading your repo, analyzing diffs, and posting findings directly to pull requests. It supports Anthropic API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry for authentication.

Why it's on the list

• Anthropic's GitHub Action that runs Claude Code on pull requests and issues when someone mentions @claude.
• Fits teams that already run CI on GitHub and want Claude to review or implement changes in the pipeline. You supply your own Anthropic API access.
• The action is free and MIT licensed; the model usage it triggers is billed to your account.
• Proof: 9,872 GitHub stars, read from the GitHub API for anthropics/claude-code-action on 2026-11-10, published by Anthropic.

Key features

• Official Anthropic-maintained GitHub Action
• Responds to @claude mentions in PR comments and issue assignments
• Implements code changes and opens PRs autonomously
• Multi-provider auth: Anthropic API, Bedrock, Vertex AI, Foundry
• Configurable triggers: PR comments, issue assignments, push events, manual dispatch

Best for

• Teams that want Claude Code reviewing every PR automatically
• Organizations needing AI-assisted code review and implementation in CI/CD
• Developers who want to delegate routine PR work (docs updates, test writing, bug fixes) to Claude

Pricing

• Free and open source
• You pay only for Claude API usage

Pros

• Official Anthropic support means reliability and long-term maintenance
• Multi-provider auth gives flexibility for enterprise deployments with existing cloud contracts
• Integrates naturally with existing GitHub workflows without changing developer habits

Cons

• API costs can add up on high-volume repos with many PRs
• Limited to GitHub (no GitLab or Bitbucket support)
• Requires careful prompt engineering in workflow configs to get consistent results

6. github-mcp-server, best official MCP server for GitHub workflows

github-mcp-server is GitHub's official MCP (Model Context Protocol) server that gives Claude Code native access to repositories, pull requests, issues, branches, Actions workflows, and code search. Instead of Claude Code guessing about your repo state, it can query GitHub's API directly through a standardized protocol.

This is the most practical MCP server for any developer who works with GitHub daily. Compatible with VS Code 1.101+, Claude Desktop, Cursor, and Windsurf.

Why it's on the list

• GitHub's official MCP server. It lets Claude Code read repositories, issues and pull requests and work with Actions through GitHub's API.
• Fits any developer or team that keeps work on GitHub and wants Claude to query it directly.
• Free and open source under the MIT license.
• Proof: 34,361 GitHub stars, read from the GitHub API for github/github-mcp-server on 2026-11-10, published by GitHub.

Key features

• Official GitHub-maintained MCP server
• Full access to repos, PRs, issues, branches, commits, and Actions
• Code search across your entire organization
• Compatible with Claude Code, VS Code, Claude Desktop, Cursor, Windsurf
• Supports GitHub Enterprise and fine-grained personal access tokens

Best for

• Any developer using Claude Code with GitHub repositories (which is almost everyone)
• Teams that want Claude Code to create PRs, manage issues, and trigger CI workflows
• Organizations on GitHub Enterprise that need custom workflow automation with proper access controls

Pricing

• Free and open source
• Requires a GitHub personal access token (free to create)

Pros

• Official GitHub support guarantees compatibility and security
• Turns Claude Code into a full GitHub collaborator instead of a code generator
• Works across multiple AI coding tools, not locked to Claude Code

Cons

• Requires initial MCP configuration setup (not zero-config)
• Token permissions need careful scoping for team environments
• Some advanced GitHub features (Projects, Discussions) have limited MCP coverage

7. awesome-claude-code-toolkit, best all-in-one skills and agent collection

awesome-claude-code-toolkit is a large pre-packaged toolkit for Claude Code, bundling 135 agents, 35 curated skills (plus 400,000+ via SkillKit), 42 commands, 176+ plugins, 20 hooks, 15 rules, 7 templates and 14 MCP configs (repo description, 2026-10-21) into a single installable package.

Where awesome-claude-code (entry #8) is a directory of links, this toolkit is a working installation. You clone it and run it. The SkillKit integration is particularly useful: it connects to a marketplace of over 400,000 community skills, searchable from within Claude Code.

Why it's on the list

• A pre-packaged toolkit of 135 agents, 35 curated skills, 42 commands, 176+ plugins and 20 hooks, per the repo description.
• Fits developers who want a large starter setup in one install and will prune what they do not use.
• Free under the Apache-2.0 license.
• Proof: 2,829 GitHub stars, read from the GitHub API for rohitg00/awesome-claude-code-toolkit on 2026-11-10, author rohitg00; the last push was 2026-05-12, so treat it as unmaintained until that changes.

Key features

• 135 pre-built agents for specialized tasks
• SkillKit integration: access 400,000+ community skills from within Claude Code
• 42 ready-to-use commands and 20 hooks
• 14 MCP configs pre-wired
• 7 project templates for common setups

Best for

• Developers who want a large collection of agents and skills without curating their own
• Teams exploring what's possible with Claude Code and want maximum surface area
• Anyone who wants SkillKit marketplace access integrated into their setup

Pricing

• Free and open source
• SkillKit marketplace may have paid skills from third parties

Pros

• Agents, skills, commands and configs in one clone
• SkillKit integration provides a practically unlimited skill library
• Well-organized with clear categories and documentation

Cons

• Quantity over quality: not every skill or agent is production-grade
• Large installation footprint may include tools you never use
• SkillKit dependency means some skills rely on external marketplace availability

8. awesome-claude-code, best curated directory of plugins and resources

awesome-claude-code is a large curated list of Claude Code skills, hooks, slash commands, agent orchestrators, applications, and plugins. With 57,668 stars (2026-11-10), it is the community's central index for discovering Claude Code extensions.

This is not a toolkit you install. It's a reference directory organized by category, linking to individual repos with descriptions and quality indicators. Think of it as the "awesome list" pattern applied to Claude Code.

Why it's on the list

• A hand-picked list of Claude Code skills, hooks, slash commands, orchestrators, applications and plugins.
• Fits anyone browsing the ecosystem, at any skill level. It is a directory, not something you install.
• Free to read. The repo uses a custom license (CC BY-NC-ND 4.0), so you cannot reuse the list commercially.
• Proof: 57,668 GitHub stars, read from the GitHub API for hesreallyhim/awesome-claude-code on 2026-11-10, author hesreallyhim.

Key features

• Categorized directory: skills, hooks, slash commands, MCP servers, IDE integrations
• Quality indicators and community ratings for listed repos
• Regular updates tracking the fast-moving Claude Code ecosystem
• Contribution guidelines ensure listed repos meet quality standards
• Cross-references related repos within categories

Best for

• Developers browsing for specific skills or plugins for a particular use case
• Anyone new to the Claude Code ecosystem who wants a structured overview
• Contributors looking to promote their own Claude Code tools to a large audience

Pricing

• Free (it's a directory, not software), published under CC BY-NC-ND 4.0

Pros

• The widest single-page overview of the Claude Code ecosystem
• Saves hours of searching GitHub Topics and forums
• Community-driven curation filters out abandoned or low-quality repos

Cons

• Not installable: you still need to visit, evaluate, and install individual repos yourself
• Can lag behind newly released repos (curation takes time)
• No quality guarantee for linked repos beyond basic community vetting

9. claude-code-mcp, best agent-in-agent MCP server

claude-code-mcp lets you run Claude Code itself as an MCP server, the "agent-in-agent" pattern. Your primary Claude Code session can spawn secondary Claude Code instances to handle subtasks: research, code generation, testing, or file analysis, all running in parallel with permissions automatically handled.

This is an advanced tool for developers who want to orchestrate multiple Claude Code agents from a single session. It's particularly powerful for complex multi-step workflows where different subtasks benefit from isolated contexts.

Status check, 2026-11-10: GitHub shows this repository as archived by its owner, which makes it read-only. It still documents the agent-in-agent pattern well, but expect no new fixes. Because it runs Claude Code with permissions bypassed, review that setting before using it anywhere near production. Our Claude Code security audit covers shell and MCP permissions like this one.

Why it's on the list

• Runs Claude Code as a one-shot MCP server, so one agent can call another.
• Fits developers experimenting with agent-in-agent setups.
• Free under the MIT license. The repo is archived, so it gets no new updates.
• Proof: 1,315 GitHub stars, read from the GitHub API for steipete/claude-code-mcp on 2026-11-10, author steipete.

Key features

• Runs Claude Code as a one-shot MCP server
• Automatic permission bypassing for headless execution
• Parallel agent execution from a single session
• Configurable context isolation between agent instances
• Works with any MCP-compatible host (Claude Code, Claude Desktop, Cursor)

Best for

• Advanced developers building multi-agent workflows in Claude Code
• Teams orchestrating complex tasks that benefit from parallel processing
• Anyone who wants to use Claude Code to control other Claude Code instances

Pricing

• Free and open source
• API costs scale with the number of agent instances spawned

Pros

• Gives you genuine multi-agent orchestration without external frameworks
• Clean MCP integration means it works within Claude Code's native tool system
• Permission handling removes the friction of running headless agent instances

Cons

• API costs multiply with each spawned agent (can get expensive fast)
• Debugging nested agent interactions is harder than single-session debugging
• Niche use case: most developers won't need agent-in-agent orchestration

10. claude-context, best semantic code search for large codebases

claude-context is an MCP plugin that adds semantic code search to Claude Code. Instead of relying on Claude Code's built-in file reading (which processes files sequentially), claude-context indexes your entire codebase and lets Claude Code query it semantically. It finds relevant code by meaning rather than exact string matching.

For large codebases where Claude Code's default context window isn't enough, this is the fix. It gives Claude Code deep awareness of your entire project without hitting token limits. For the underlying discipline of managing what actually goes into that window, see our context engineering guide.

Why it's on the list

• An MCP plugin that indexes a codebase and lets Claude Code search it by meaning rather than exact strings.
• Fits developers working in large codebases where plain file reading falls short.
• Free under the MIT license; indexing needs a vector database and an embedding provider, which can carry their own costs.
• Proof: 12,683 GitHub stars, read from the GitHub API for zilliztech/claude-context on 2026-11-10, published by Zilliz.

Key features

• Semantic code search across entire codebases using vector embeddings
• MCP integration: works natively within Claude Code's tool system
• Indexes code on first run, then updates incrementally
• Finds relevant code by meaning even when keywords don't match
• Supports multiple programming languages and file types

Best for

• Teams working on large monorepos or codebases with 100K+ lines of code
• Developers who find Claude Code loses context on big projects
• Anyone who needs Claude Code to understand cross-file relationships in complex architectures

Pricing

• Free and open source
• Requires local compute for indexing (no cloud costs)

Pros

• Dramatically improves Claude Code's understanding of large codebases
• Semantic search finds relevant code that keyword search would miss
• Incremental indexing means fast updates after initial setup

Cons

• Initial indexing takes time on very large codebases
• Requires local resources (RAM, disk) for the vector index
• Adds complexity to the Claude Code setup compared to using it standalone

11. claude-code-ultimate-guide, best learning resource and template library

claude-code-ultimate-guide is a massive documentation project covering Claude Code from beginner to power user. Its repo description claims 430K+ lines across the guide, runnable examples, templates and a knowledge quiz (checked 2026-10-21).

This is not a tool or framework. It's a reference library. If you're serious about mastering Claude Code (or onboarding your team), this is where you start.

Why it's on the list

• A documentation repo with a guide, runnable examples, templates and a quiz covering Claude Code from beginner to advanced use.
• Fits newcomers learning Claude Code and teams that want templates to copy.
• Free under the CC BY-SA 4.0 license.
• Proof: 6,400 GitHub stars, read from the GitHub API for FlorianBruniaux/claude-code-ultimate-guide on 2026-11-10, author Florian Bruniaux.

Key features

• Guide, runnable examples (agents, skills, hooks, workflows), starter CLAUDE.md templates and a knowledge quiz, per the README

Best for

• Developers new to Claude Code who want a structured learning path
• Team leads onboarding their engineering team onto Claude Code
• Anyone who wants copy-paste templates for common AI automation patterns

Pricing

• Free. Guide text under CC BY-SA 4.0, reusable templates under CC0 1.0 (README)

Pros

• Covers setup through advanced workflows in one repo
• Quiz questions make it useful for team training and certification
• Templates are production-ready, no toy examples

Cons

• Documentation-heavy: not something you install and run
• Can become outdated as Claude Code evolves rapidly (check last commit date)
• Overwhelming volume for developers who just want quick answers

12. ccusage, best for tracking Claude Code token usage and cost

ccusage is a command-line tool that reads the usage data Claude Code already stores on your machine and turns it into daily, weekly, monthly and per-session reports of tokens and estimated cost. You run it through npx or bunx without a global install. On 2026-11-10 the repo showed 19,389 stars, an MIT license and a latest release of v20.0.48, published 2026-11-02.

Most other repos here add capability. This one shows what that capability costs you. Its blocks report tracks usage inside Claude's 5-hour billing windows, which is where most heavy users hit limits first. For more terminal utilities like it, see our best Claude Code CLI tools roundup.

Why it's on the list

• A command-line tool that reads Claude Code's local usage logs and reports tokens and estimated cost by day, week, month and session.
• Fits anyone on Claude Code who wants to see where token spend goes. Run it with npx or bunx.
• Free under the MIT license.
• Proof: 19,389 GitHub stars, read from the GitHub API for ccusage/ccusage on 2026-11-10, author ryoppippi.

Key features

• Daily, weekly, monthly and session reports, with --json export
• ccusage blocks for Claude's 5-hour billing windows, with active block monitoring
• A status line integration for Claude Code hooks (marked beta in the README)
• Per-project grouping with --instances and an --offline mode that uses cached pricing data
• Also reads Codex, OpenCode, Gemini CLI, GitHub Copilot CLI and other agent CLIs, 18 sources in the README table

Best for

• Developers on usage-capped plans who want to see which sessions burn the most tokens
• Team leads who want a per-project JSON export to compare spend across repos

Pricing

• Free, MIT license

Pros

• No setup and no API key: it works from local files
• One report covers Claude Code plus other coding agents you run on the same machine
• Frequent releases: v20.0.48 shipped on 2026-11-02, ten patch releases after the v20.0.38 we saw on 2026-10-21

Cons

• Dollar figures come from the LiteLLM pricing file the tool ships with, not from your Anthropic invoice, so treat them as estimates
• It only sees usage on the machine where it runs, so team totals need each developer to export their own report
• The README's quick start pulls the latest version. Pin a version in shared team scripts

Skip it if

• Your Claude usage runs through the API from your own application code. ccusage reads agent CLI logs, not API traffic from your services

13. claude-plugins-official, best official source for Claude Code plugins

anthropics/claude-plugins-official is the repo behind Claude Code's built-in plugin marketplace. Its marketplace.json listed 317 plugins on 2026-11-10, up from 315 on 2026-10-14. Internal plugins from Anthropic sit in /plugins and partner and community plugins in /external_plugins. The two newest entries, Apple's Xcode plugin and Anthropic's champion-kit, were added on 2026-11-06.

You rarely clone it, because Claude Code adds this marketplace on first launch. Its value on GitHub is reading the source of a plugin before you run /plugin install name@claude-plugins-official. The README warns that Anthropic does not control what MCP servers or files a plugin includes. For the plugins worth installing from it, see our best Claude Code plugins guide.

Why it's on the list

• Anthropic's official, Anthropic-managed directory of Claude Code plugins, and the source for /plugin > Discover.
• Fits anyone installing plugins who wants to read the code first.
• Free; the repo is Apache-2.0 licensed, and each external plugin carries its own license.
• Proof: 46,697 GitHub stars, read from the GitHub API for anthropics/claude-plugins-official on 2026-11-10, published by Anthropic.

Pros

• One place to audit every official plugin's skills, hooks and MCP config
• Since 2026-10-26 it is also where Anthropic's older demo marketplace installs from

Cons

• Listing here is not a security review of third-party plugins, per the README
• 317 entries is a lot to browse without a curated shortlist

14. tstack, best skills for shipping pull requests

tstack is Theo Browne's set of the agent skills he says he uses every day, published on 2026-10-24 and at 3,422 stars by 2026-11-10. It holds nine skills for Claude Code and Codex, all about getting a change merged: full-send proves, reviews, files, babysits and merges a change; council gets review from three models from different providers until they agree; babysit-pr watches CI and review bots and fixes real findings; takeover finishes someone else's PR and keeps their credit.

Why it's on the list

• Nine PR-workflow skills, installed by copying skills/* into ~/.claude/skills/.
• Fits developers who already live in GitHub PRs and want the agent to carry a change to merge.
• Free under the MIT license. Needs an authenticated GitHub CLI, and council and sol-loop need a harness that can run other providers' models.
• Proof: 3,422 GitHub stars, read from the GitHub API for t3dotgg/tstack on 2026-11-10.

Pros

• Small and readable: nine skills, with the dependencies between them listed in the README
• Works in Claude Code and Codex

Cons

• Some skills use T3 Code features and fall back to native tools without them
• Last push was 2026-10-28, so it is too new to judge long-term upkeep

15. Floorplan, best for mapping an unfamiliar codebase

Floorplan is a Claude Code and Codex plugin, created on 2026-10-27, that reads your codebase and draws an interactive map. Every part, file and web request gets a plain-English name, and every connection is drawn, including requests between a frontend and a backend in another repo. It had 3,404 stars on 2026-11-10, version 0.7.1 and an MIT license.

Install it from its own marketplace with claude plugin marketplace add haydenschmitty/floorplan, then claude plugin install floorplan@floorplan. It runs on your existing Claude Code or Codex plan with no API key, per the README.

Why it's on the list

• A plugin that turns a codebase into a zoomable map with a health check and a view of what your agent just changed.
• Fits builders who shipped an app with AI help and lost track of how it is wired, and teams onboarding someone new.
• Free under the MIT license; needs Node.js 18+.
• Proof: 3,404 GitHub stars, read from the GitHub API for haydenschmitty/floorplan on 2026-11-10.

Pros

• The health check flags keys in code, open Supabase tables and Firebase rules, and fakeable Stripe webhooks, per the README
• Active: the last push was on 2026-11-10

Cons

• Two weeks old, so expect frequent changes
• Not in Anthropic's official marketplace, so read the source before you install

How to choose the best Claude Code GitHub repo

1) Do you need skills, a framework, or infrastructure?

If you want plug-and-play capabilities you can use immediately, start with a skills repo. Ay-Skills gives you 10 diverse skills. Superpowers gives you a complete TDD methodology. For infrastructure (CI/CD, GitHub integration, code search), look at claude-code-action, github-mcp-server, and claude-context.

• If you want skills: start with Ay-Skills or Superpowers, and keep anthropics/skills open as the reference format. Add tstack if most of your work ends in a pull request
• If you want infrastructure: start with github-mcp-server and claude-code-action, then add ccusage to watch token spend
• If you want everything: start with Everything Claude Code or awesome-claude-code-toolkit

2) How experienced is your team with Claude Code?

For teams just getting started, claude-code-ultimate-guide provides the learning foundation, and awesome-claude-code gives you a map of the ecosystem. For experienced teams, Superpowers and Everything Claude Code provide the kind of structured workflows and agent orchestration that justify their complexity.

• If your team is new to Claude Code: start with the ultimate guide and awesome-claude-code
• If your team is experienced: go straight to Superpowers or ECC
• If you need to hire developers who already know the Claude Code ecosystem, engineer placement services can connect you with pre-vetted engineers familiar with these tools
• If the agents you build need to read from and write to your CRM, ERP or support desk, AI integration services cover the connectors, scoped credentials and logged writes

3) What's your risk tolerance for dependencies?

Official repos (claude-code-action, anthropics/skills, github-mcp-server) have institutional backing and will be maintained long-term. Community repos with massive star counts (Superpowers, ECC) have strong community support but could shift direction. Smaller repos (Ay-Skills, claude-context) are more focused but depend on individual maintainers. Check the archived flag and the last push too: claude-code-mcp is still archived on 2026-11-10, and awesome-claude-code-toolkit has not been pushed since 2026-05-12. Whichever you pick, check what its hooks and MCP servers are allowed to run before you install them; our Claude Code development agency hardens that environment for teams.

• If you need stability: prioritize official Anthropic and GitHub repos
• If you want the newest features: community repos move faster than official ones
• If you want quality over quantity: Ay-Skills and Superpowers are opinionated

If you're building custom Claude Code skills or workflows for your team and want expert guidance, AY Automate's AI agent development team specializes in building production-grade AI tooling. Book a free discovery call to discuss your Claude Code setup.

FAQ

What are the best GitHub repos for Claude Code in 2026?

The top repos are Ay-Skills (10 plug-and-play production skills), Superpowers (TDD-enforced dev framework, 315K+ stars on 2026-11-10), Everything Claude Code (complete agent harness, 300K+ stars), anthropics/skills (Anthropic's official skills) and claude-code-action (Anthropic's official GitHub Actions integration). The best choice depends on whether you need skills, a framework, or CI/CD integration. The newest repos worth watching are tstack and Floorplan, both created in late October 2026.

What is a Claude Code skill?

A Claude Code skill is a SKILL.md file (with optional supporting scripts and assets) that you place in your project's .claude/skills/ directory. Claude Code auto-loads it and gains the capabilities described in the skill file. Skills can add slash commands, enforce development methodologies, connect to external services, or automate complex workflows.

Are Claude Code GitHub repos free to use?

Yes, all 15 repos in this list are free to use, but licenses differ. Most are MIT or Apache 2.0. awesome-claude-code uses CC BY-NC-ND 4.0, the ultimate guide uses CC BY-SA 4.0, Anthropic's document skills are source-available, and Ay-Skills is MIT. Your real cost is Claude usage, which claude-code-mcp can multiply by spawning parallel agents.

How do I install a Claude Code skill from GitHub?

Clone or download the repo, then copy the skill folder into your project's .claude/skills/ directory. Claude Code automatically detects and loads any SKILL.md file in that directory. No additional configuration is needed for most skills.

What is an MCP server for Claude Code?

MCP (Model Context Protocol) is the standard that lets Claude Code connect to external tools and data sources. An MCP server exposes capabilities (like GitHub API access or semantic code search) through a standardized interface. You configure MCP servers in your Claude Code settings, and Claude Code can then use their tools during conversations. For the servers worth installing first, see the 12 best Claude Code MCP servers.

Which Claude Code GitHub repo should I install first?

Start with github-mcp-server if you work with GitHub (most developers do), as it gives Claude Code native access to your repos and PRs. Then add Ay-Skills for a curated set of production skills. If you want to enforce better development practices, add Superpowers for TDD methodology.

How do I get my team up to speed on Claude Code tools?

Start with claude-code-ultimate-guide for self-paced learning, then have your team experiment with one or two repos from this list. For a faster ramp-up, custom AI training programs can teach your team how to evaluate, install, and build on these tools effectively. For Claude Code-specific onboarding that covers setup, CLAUDE.md conventions, and hands-on workflow sessions on your actual codebase, our Claude Code training workshops run in one to three days.

Does Anthropic have an official Claude Code skills repository?

Yes. anthropics/skills is Anthropic's public repository for Agent Skills. It contains example skills, the document skills behind Claude's Word, PDF, PowerPoint and Excel output, the Agent Skills spec and a template. In Claude Code you add it with /plugin marketplace add anthropics/skills and then install document-skills or example-skills.

How do I track Claude Code token usage and cost?

Run ccusage, an MIT-licensed CLI that reads the usage data Claude Code stores locally. It prints daily, weekly, monthly and per-session reports, plus a blocks view for Claude's 5-hour billing windows. The cost column is an estimate from a public pricing file, so check your invoice or plan dashboard for the billed amount.

Where is Claude Code on GitHub, and how do I connect it to my repos?

Claude Code's own GitHub repository is anthropics/claude-code (155,154 stars on 2026-11-10), which holds the issue tracker and the changelog. Its demo plugin marketplace now installs plugins from anthropics/claude-plugins-official. To connect Claude Code to your GitHub repos, run /install-github-app inside Claude Code: it installs the Claude GitHub App, adds the authentication secret and prepares a pull request with a claude-code-action workflow. Anthropic's GitHub Actions docs say this setup works only with github.com repositories.

Which Claude Code GitHub repos are official?

Four repos on this list come from the platform owners. Anthropic maintains claude-code-action for GitHub Actions, anthropics/skills for Agent Skills and claude-plugins-official for plugins. GitHub maintains github-mcp-server. The other eleven are community projects, including Superpowers and Everything Claude Code, and Ay-Skills, which is maintained by AY Automate, the team behind this site.

Is claude-code-mcp still maintained?

No new work is expected. On 2026-11-10 GitHub still showed steipete/claude-code-mcp as archived by its owner, which makes the repository read-only. The code still runs and still explains the agent-in-agent pattern, but bugs will not be fixed there. If you depend on it, fork it or keep a pinned copy you can patch yourself.

Do these repos work with Claude Fable 5.2?

Yes. Skills, plugins, MCP servers and hooks do not depend on the model. Claude Code 2.1.321, released 2026-11-09, made Claude Fable 5.2 the default Fable model, and every repo here runs the same way on it. Cost is the part that changes: Fable 5.2 input is $8 per million tokens, so large skill packs and harnesses that load many descriptions per turn cost more per session. ccusage (#12) shows where that spend goes.

Can I use multiple Claude Code repos together?

Yes. Skills, MCP servers, and frameworks are designed to coexist. You can run Superpowers for development methodology, github-mcp-server for GitHub access, and Ay-Skills for specialized capabilities all in the same project. The only constraint is that conflicting skills (two skills that handle the same trigger differently) may need manual resolution.

What is the difference between Superpowers and Everything Claude Code?

Superpowers focuses on development methodology: it enforces TDD, structured planning, and code review practices. Everything Claude Code is a full agent harness: it includes agents, security scanning, memory optimization, model routing, and cross-platform support. Superpowers makes you a better developer. ECC gives you more infrastructure. Many teams use both.

Once you have found the repos that fit your stack, see the best Claude Code project templates guide for pre-configured starters that wire these repos into a working codebase with CLAUDE.md, agents, hooks, and MCP servers already set up.

If you would rather have these workflows built and run for you instead of wiring the repos together yourself, that is what an AI automation agency does end to end.

Running several of these repos concurrently at scale also means watching your usage against Claude Code rate limits.