Everything Claude Code: The Complete Resource Guide for 2026

Claude Code authored roughly 12% of all public GitHub commits by mid-May 2026, up from 4% just three months earlier. That 42,896x increase from its research preview in early 2025 made it the fastest-adopted developer tool in recent memory, with 4.2 million weekly active developers averaging 20 hours per week with the tool.

Yet most developers still discover Claude Code features one at a time. They stumble across skills one week, learn about MCP servers the next, and only hear about routines from a coworker months later. No single page catalogs the full ecosystem, the community projects, and the plugin marketplaces that have grown around it.

This guide fixes that. Below is every major capability organized by category, followed by deep dives into surfaces, extensions, orchestration, and the community ecosystem.

Claude Code ships across six primary surfaces. Each connects to the same underlying engine, so your CLAUDE.md files, settings, and MCP servers work identically regardless of where you open a session.

Terminal CLI

The original surface and still the most capable. Install with a single command on Mac or Linux (curl -fsSL https://claude.ai/install.sh | bash), or through Homebrew, WinGet, apt, dnf, and apk on other platforms. The CLI supports piping, scripting, and composition with standard Unix tools. Feed it log output, chain it with git commands, or run it headless in CI pipelines. On Windows, a native installer and PowerShell one-liner are also available.

VS Code and Cursor

The VS Code extension adds inline diff review, @-mentions for files and symbols, plan approval workflows, and conversation history directly in the editor. Search "Claude Code" in the Extensions view or install from the marketplace. The same extension works in Cursor without modification.

JetBrains

A plugin for IntelliJ IDEA, PyCharm, WebStorm, and other JetBrains IDEs. It provides interactive diff viewing and selection context sharing. Install from the JetBrains Marketplace. The plugin requires the terminal CLI installed separately, and once connected the two stay in sync.

Desktop App

A standalone application for Mac and Windows. Review diffs visually, run sessions side by side, schedule recurring tasks, and kick off cloud sessions without touching the terminal. Download from claude.ai for Intel and Apple Silicon Macs, plus x64 and ARM64 Windows. Sessions started in the desktop app can be continued on other surfaces.

Web

Run Claude Code at claude.ai/code with no local setup. Start long-running tasks from any browser and check back when they finish. Work on repos you don't have cloned locally. Move a web session into your local terminal with claude --teleport when you need full CLI access.

iOS

The Claude iOS app supports full Claude Code sessions. Start a task from your phone while commuting, then pull it into your terminal or desktop app when you sit down. Combined with Remote Control, you can continue any local session from your phone without losing context.

Beyond these six surfaces, Claude Code integrates directly with Slack (mention @Claude to get pull requests from bug reports), GitHub Actions and GitLab CI/CD for workflow automation, and Chrome for debugging live web applications.

Claude Code becomes more useful once you customize it. Four primary extension types let you add persistent knowledge, automate lifecycle events, connect external services, and package everything for distribution.

Skills

A skill is a markdown file containing instructions, workflows, or domain knowledge that Claude Code loads on demand. Drop a SKILL.md file into your project's .claude/skills/ directory and invoke it with a slash command. A /deploy skill might contain your team's deployment checklist. A /review-component skill might define how your team evaluates React components. Skills replace the habit of pasting the same long prompt every session.

Skills can also load automatically when Claude detects relevant context. If you have a skill for database migrations, Claude surfaces it when it encounters migration-related files without you asking.

Hooks

Hooks fire shell commands at lifecycle events: before or after tool calls, on session start, on session end. Think of them as git hooks for Claude Code. Common uses include auto-formatting after every file edit, running lint checks before a commit, posting to Slack when a session finishes, or logging tool usage to a local file. Hooks are deterministic and run without prompting.

MCP Servers

The Model Context Protocol connects Claude Code to external services and data sources. Configure MCP servers in your project settings, and Claude Code can read design docs from Google Drive, update tickets in Jira, pull messages from Slack, query PostgreSQL databases, automate browsers, or interact with any service that exposes an MCP endpoint.

Fast.io, for example, exposes workspace operations through Streamable HTTP at mcp.fast.io, giving Claude Code direct access to file storage, AI-powered semantic search, and workflow tools through a single MCP connection. The MCP skill documentation covers the full tool surface. Other popular MCP servers cover GitHub, Notion, Linear, Sentry, and browser automation frameworks.

Plugins

A plugin bundles skills, hooks, sub-agents, and MCP servers into one installable unit. Plugin skills are namespaced (/my-plugin:review) so different plugins coexist without naming conflicts. The official Anthropic plugin directory ships pre-installed with Claude Code, and community marketplaces offer thousands more. You can build your own plugins and distribute them to your team or publish them publicly.

CLAUDE.md and Auto-Memory

CLAUDE.md is a markdown file at your project root that Claude Code reads at the start of every session. Use it for coding standards, architecture decisions, preferred libraries, or review checklists. Auto-memory complements this by saving things Claude learns while working. Build commands, debugging techniques, and environment quirks get persisted automatically and injected into future sessions without you writing anything.

Claude Code is not limited to single-agent sessions. Its orchestration features let you run agents in parallel, coordinate teams, schedule recurring work, and push events from messaging platforms into live sessions.

Sub-Agents

When a task gets complex, Claude Code spawns sub-agents with their own context windows, system prompts, tool permissions, and optionally different models. A lead agent coordinates the work, assigns subtasks, collects results, and merges them. Sub-agents keep the main session focused while handling specialized work in isolation. For example, one sub-agent might refactor a module while another writes tests for it, both reporting back to the lead.

Agent SDK

The Agent SDK lets you build custom agents powered by Claude Code's tools and capabilities. You control orchestration, tool access, permissions, and output format. This is the path for building production-grade agents: automated code review bots, documentation generators, multi-repo migration tools, or anything that needs programmatic control beyond what interactive sessions provide.

Background Agents

Run several full Claude Code sessions in parallel and monitor them from a single screen. Background agents are useful when work spans repositories or requires independent workstreams that converge later. The agent view provides live status, logs, and the ability to intervene in any running session.

Routines

Routines run Claude Code on a schedule using Anthropic-managed infrastructure. They execute even when your computer is off and can trigger on API calls or GitHub events. Common uses: morning PR reviews, overnight CI failure analysis, weekly dependency audits, post-merge documentation syncing. Create routines from the web, the desktop app, or with /schedule in the terminal.

Channels

Channels push events from external messaging platforms into Claude Code sessions. Telegram, Discord, iMessage, Slack, WhatsApp, Signal, email, and custom webhooks can all feed into a running session. This turns Claude Code into a reactive agent that responds to messages, alerts, and external triggers rather than waiting for terminal input.

Remote Control and Teleport

Remote Control lets you continue a local session from your phone or any browser. Teleport moves a web or iOS session into your local terminal with claude --teleport. You can start work on any surface and finish it on any other. Dispatch, accessible from the desktop app, lets you message a task from your phone and open the session it creates when you get back to your desk.

For teams building Claude Cowork patterns where agents and humans share the same project context, these orchestration features pair naturally with shared workspaces. Fast.io provides the storage and intelligence layer where both agents and humans access the same files, search index, and collaboration tools through a single workspace.

Give your Claude Code agents a workspace that persists

Storage with an MCP endpoint at mcp.fast.io for reads, writes, and semantic search. Transfer ownership of anything an agent builds to your team. Starts with a 14-day free trial.

Start 14-Day Trial

The community around Claude Code has built an ecosystem of configuration frameworks, curated lists, and plugin directories that rival the official feature set in breadth.

Everything Claude Code (ECC)

The largest community project, Everything Claude Code reached 100,000 GitHub stars and 13,000 forks by mid-2026. Created by Affaan Mustafa after winning an Anthropic hackathon in September 2025, ECC ships 28 specialized agents, 119 skills, and 60 commands that work across Claude Code, Cursor, Codex, and OpenCode from a single repository. Version 1.9.0 introduced selective-install architecture, six new agents, support for 12 language ecosystems, and AgentShield, a security scanner with 1,282 tests and 102 static analysis rules.

With 113 contributors and translations in seven languages, ECC has become something like the .editorconfig standard for AI coding tools. Browse it at github.com/affaan-m/ecc.

Awesome Claude Code Toolkit

The awesome-claude-code-toolkit repository catalogs 135 agents, 35 curated skills, 42 commands, 176+ plugins, 20 hooks, 14 MCP configurations, 26 companion apps, 7 templates, and 52 ecosystem entries. It connects to the SkillKit marketplace for access to over 400,000 community-contributed skills.

Other Notable Collections

• claude-skills: 337 skills across engineering, marketing, product, compliance, and operations domains, with 30+ agents and 70 custom commands
• claude-code-plugins-plus-skills: 425 plugins, 2,810 skills, and 200 agents with an open-source marketplace and its own CLI package manager (ccpi)
• awesome-claude-plugins: Tracks real adoption metrics across GitHub repositories, measuring which plugins see production use versus those that collect stars without traction

Plugin Marketplaces

Several directories aggregate Claude Code extensions at scale:

• The official Anthropic plugin directory (claude-plugins-official) ships pre-installed with every Claude Code installation
• Community marketplace directories index over 21,600 skills, 2,500 marketplace entries, and 12,500 MCP server configurations
• Specialized directories focus on specific domains like security auditing, frontend development, or DevOps automation

The sheer volume means curation matters more than quantity. Start with the official directory and ECC, then branch out to domain-specific marketplaces as your needs narrow.

Claude Code agents write code, create files, build documentation, and generate project scaffolds. Everything they produce needs to live somewhere, particularly when agents collaborate or when output needs to reach a human.

Local Filesystem

The default approach. Claude Code reads and writes files in your project directory. Simple, fast, and fine for single-developer workflows where output stays on one machine. The limitation appears when agents need to share files across sessions, hand deliverables to clients, or maintain persistent state between runs.

Object Storage (S3, GCS)

For raw file persistence without collaboration features. Agents can write to S3 or GCS buckets through MCP servers or direct API calls. You get durability and scale but no built-in search, no versioning UI, and no sharing workflows designed for handoff.

Consumer Cloud (Google Drive, Dropbox)

Better for basic file sharing but not built for agent workflows. These services lack agent-specific free tiers, built-in semantic search across stored documents, and native MCP endpoints that would let agents interact without workarounds.

Fast.io

Fast.io treats agents as first-class workspace participants. Connect through the MCP server at /mcp and Claude Code agents get read/write workspace access, semantic search through Intelligence Mode (built-in RAG with citations, no separate vector database needed), file versioning, granular permissions, and branded sharing workflows for client delivery.

Ownership transfer lets an agent build a workspace, populate it with files, and hand ownership to a human while keeping admin access for future updates. This makes the agent-to-human handoff a clean transition rather than a folder dump. Agents and humans share the same intelligence layer, so a question asked in the UI searches the same indexed content that agents wrote to via MCP.

Plans start with a 14-day free trial, with workspaces, monthly credits, shares, and storage scaled to each tier. For Claude Code workflows that produce deliverables for clients or colleagues, a workspace accessible to both agents and humans closes the gap between "agent finished" and "team can use it."