Three MCPs that cover the majority of typical Claude Code workflows. Start here, add everything else later.
The "yellow pages" — master directories that point at everything else. Start here if you're lost.
The hand-curated, canonical master index of skills, hooks, slash commands, agent orchestrators, apps, and plugins. The one list to rule them all.
A second curated catalog focusing on IDE integrations and frameworks. Good cross-reference when the main list isn't giving you what you need.
Specifically focused on the plugin system — the official extension format that bundles commands, agents, hooks, and MCPs together into one installable unit.
A robot-built list that tracks which plugins are actually being adopted in real repos. Useful for "what's actually popular right now" rather than what someone thought was cool.
100+ specialized "subagents" — mini-Claudes pre-loaded with personalities like "security reviewer," "test writer," or "docs generator."
Tools that prevent Claude from forgetting what it's doing, losing the plot mid-session, or misunderstanding your codebase.
Plugs Claude into real language servers (LSP) so it understands code the way an IDE does — jump-to-definition, find-all-references, rename symbols. Best for large codebases.
Stores your whole codebase in a vector database and only feeds Claude the relevant chunks on demand, saving tokens and money on large projects.
Recaps each session, compresses the history, and re-injects relevant memory on next run. Claude remembers decisions made in previous sessions.
You don't have to live in the terminal. These wrappers bring Claude Code into desktop apps, mobile, and your existing IDE.
A desktop app (Tauri/React) wrapping Claude Code with session history, checkpoints/timeline, custom agent designer, MCP manager, and usage dashboard.
Crystal is now Nimbalyst. Runs multiple Claude Code sessions side-by-side in separate git worktrees so you can try three approaches to the same task simultaneously.
Mobile + web client for Claude Code with end-to-end encryption, push notifications, and voice input. Also available as happy-cli.
A web kanban board where each card is a task being worked on by a coding agent (Claude, Codex, Gemini, etc.) running in parallel.
Community plugin bringing Claude Code into JetBrains IDEs — IntelliJ, PyCharm, WebStorm. For those who don't live in VS Code.
Running many Claudes at once, coordinating them, and keeping them from stepping on each other.
Runs 20–50 Claude Code sessions in parallel with lock-based coordination and a tmux monitoring dashboard.
Lightweight terminal app juggling multiple AI agents — Claude Code, Codex, Aider, OpenCode — in separate isolated workspaces.
Enterprise orchestration platform with a "queen + worker" hive-mind topology, self-learning memory, federated comms, and 314 MCP tools. Formerly claude-flow.
Decomposes a task into a dependency graph and runs independent pieces simultaneously, visualized in a TUI.
Production-grade kit with 185 agents, 16 workflow orchestrators, and 100 commands. Pre-built Claude team rosters for common dev situations.
Plugin that hooks into plan mode to auto-decompose tasks and delegate pieces to specialized agents.
Rules that fire automatically during a Claude Code session — guardrails that don't require babysitting.
Watches what Claude is doing and blocks it from skipping tests or over-implementing. Enforces test-driven development automatically.
macOS desktop notifications when Claude finishes a task or needs your input. Clicking the notification jumps you back to the right project in VS Code.
The little info bar at the bottom of your terminal — turned into a live dashboard showing model, branch, token spend, and time until rate limit.
Highly customizable statusline with powerline support, multiple themes, model/git/token info display.
Same concept as ccstatusline but written in Rust for speed, with a TUI configurator for easy setup.
One-line rate-limit and usage bar with themes and a daemon mode for always-on monitoring.
Another themed statusline alternative. Good option if the others don't fit your terminal setup.
Token budgets are real. These tools tell you where your money went, how fast it's going, and warn you before you hit the wall.
CLI that reads local logs and reports tokens + cost by day, month, session, or 5-hour billing block. Works for Claude Code, Codex, OpenCode, Amp.
Real-time terminal monitor with ML predictions for when you'll hit your rate limit.
Hooks Node's fetch() to intercept actual API calls for the most precise cost data possible — straight from the wire.
macOS menu bar app showing live usage limits as a small icon. Glanceable at all times.
Local web dashboard for token spend and session history. More visual than CLI-only trackers.
Use cheaper models for boring work, Claude for the hard stuff. Automatic dispatching by task complexity.
Lets Claude Code talk to other models (OpenRouter, DeepSeek, Ollama, Gemini) and route tasks by type — cheap models for boilerplate, Claude for hard reasoning.
Lighter-weight variant that translates Claude's API format into OpenAI-compatible format to unlock OpenRouter's free tier.
Another lighter router variant specifically targeting OpenRouter integration for access to that ecosystem's free tier models.
Drop these in and Claude Code transforms. Big opinionated configurations that give you commands, personas, and workflows out of the box.
Drops a ~/.claude/ config with 30 specialized commands and 9 "cognitive personas" — architect, security, QA, mentor, and more.
CLI + web catalog of 600+ agents, 200+ commands, 55+ MCPs, 60+ settings, 39+ hooks, and 14 templates. The App Store for Claude Code config.
Time-saving plugin pack built by working developers — practical stuff that solves real workflow problems.
425 plugins / 2,810 skills / 200 agents bundled into an open marketplace with a ccpi CLI package manager for easy installs.
MCPs are USB ports for Claude — each one connects it to a new app or service. These are the essential ones.
The official reference MCP servers from Anthropic: filesystem, fetch, git, and other foundational tools. Boring but essential.
Full GitHub API access — read repos, open PRs, manage issues, browse code. Claude gets a full GitHub login.
Browser automation — open pages, click buttons, fill forms, scrape content. Claude gets a fully controllable web browser.
Lets Claude inspect a live Chrome browser with full DevTools access — network tab, performance panel, console, everything.
Pulls up-to-date, version-specific library docs into Claude's context on demand. Claude reads the current README instead of guessing from training data.