Skip to content

Knowledge base

Claude Code Ecosystem

Claude Code ships as a thin harness over the model, so most of its practical leverage sits in the ecosystem around its extension points: plugins, skills, hooks, MCP servers, usage trackers, routing proxies and token optimizers. The same openness makes that ecosystem large and noisy.

Written by

Updated:

Tools 4

Our position. We expect the pattern to hold: the community innovates, platforms absorb (opens in a new tab). So reach for the simplest thing that works and add an extension only when it earns its place, because a gold-plated harness becomes what slows the line.

Every extension point adds context or coordination cost. Choose each mechanism on purpose, and prune any add-on that a native feature has since replaced.

Start here

Resources

Configuration & Templates

Plugins

Skills & Workflows

  • 610ClaudeSubagents (opens in a new tab) - Claude-Flow swarm framework bundling 610+ specialized agents for software, business, analytics, and personal domains, orchestrated in parallel swarms with internet access.
  • AI Adoption Playbook (opens in a new tab) - Collection of 12 Claude Code skills for engineering leaders driving organizational AI adoption, covering fluency assessments, ROI calculation, and board-ready updates.
  • Agent Browser Skills (opens in a new tab) - Skill folder in Vercel’s agent-browser repo teaching coding agents to drive browser automation through the agent-browser CLI.
  • Anthropic Skills (opens in a new tab) - Anthropic’s official example Agent Skills covering design, development, enterprise comms, and document handling for Claude Code, Claude.ai, and the Claude API.
  • Awesome Claude Skills (opens in a new tab) - Curated list of 20+ Claude Skills for business, marketing, comms, creative media, development, and productivity, portable across Claude.ai, Claude Code, and the Claude API.
  • Awesome Skills (opens in a new tab) - GitHub org publishing standalone Claude Code skills including code review across 17+ languages, first-principles analysis, 5-whys root-cause, and a Memex session-history search engine.
  • BTP Drawio Skill (opens in a new tab) - Claude Code plugin that turns plain-English descriptions into editable SAP BTP architecture diagrams, scaffolding from official templates and validating output against SAP guidelines; MIT.
  • Better Code (opens in a new tab) - Personal collection of Claude skills aimed at writing better code and shipping more meaningful projects.
  • Build with Claude (opens in a new tab) - Marketplace and directory of Claude Code plugins bundling 175 slash commands plus subagents, hooks, and skills, installable through the /plugin command.
  • Claude Scientific Skills (opens in a new tab) - Scientific Agent Skills library covering bioinformatics, drug discovery, clinical research, and quantum computing across 37 databases, with 54 agents and 177 skills.
  • Python codebase skills (opens in a new tab) - Reusable Claude Code slash-command skills for Python codebase work: tightening type annotations, contract-style docstrings, Hypothesis property tests, mutation testing, and pre-mortem bug imagination.
  • Full-stack engineering skills (opens in a new tab) - 66 full-stack engineering skills for Claude Code spanning 12 languages, multiple frameworks, cloud, security, and ML.
  • Claude Skillz (opens in a new tab) - Grab-bag of Claude workflow utilities: multi-session task tracking, implementation analysis protocol, and an interactive launcher with custom system prompts.
  • Code Review Skill (opens in a new tab) - Modular Claude Code skill for structured code review across 17+ languages and frameworks, with 14,000+ lines of rules and checklists.
  • Color Expert Skill (opens in a new tab) - Agent skill for color science covering color spaces, accessibility, palette generation, and pigment mixing; works with Claude Code, Codex, Cursor, and Copilot.
  • Crafter Station Skills (opens in a new tab) - Crafter Station’s MIT-licensed collection of AI agent skills across context engineering, meta, music, and security, portable across Claude Code, Cursor, Copilot, and 10+ agents.
  • Fork Repository Skill (opens in a new tab) - Claude Code skill that forks the current agent into N parallel terminal windows for delegated work and context handoff on macOS and Windows.
  • Graphify (opens in a new tab) - Knowledge-graph skill that turns code, schemas, docs, and images into a queryable graph across 25 languages, with claimed ~71.5x token reduction on structural queries.
  • Hugging Face Skills (opens in a new tab) - Hugging Face’s Agent Skills covering HF CLI, model evaluation, dataset exploration, Gradio UIs, LLM fine-tuning, and vision training, compatible with Claude Code and Cursor.
  • HyperSkills (opens in a new tab) - Next.js app that auto-generates SKILL.md files for AI coding agents from web search and content extraction via the Hyperbrowser SDK.
  • Loop (opens in a new tab) - Two Claude Code skills: /prd for interactive PRD generation and /loop for atomic task execution with dependency tracking, per-step commits, and crash recovery.
  • MiniMax Skills (opens in a new tab) - MiniMax’s Agent Skills for development, document production, and media creation.
  • My Second AI Skills (opens in a new tab) - PM operating system on Claude with 70+ slash-command skills for product managers covering discovery, planning, specs, strategy, competitive analysis, and chained multi-agent workflows.
  • OK Skills (opens in a new tab) - Curated AI coding agent skills and AGENTS.md playbooks for Codex, Claude Code, Cursor, OpenClaw, and other SKILL.md-compatible tools.
  • Obsidian Skills (opens in a new tab) - Claude Skills for Obsidian that teach the agent to create and edit Markdown, Obsidian Bases (.base), and JSON Canvas (.canvas) files.
  • OpenAI Skills (opens in a new tab) - OpenAI’s official Codex skills catalog, organized into system, curated, and experimental tiers, installable by name or GitHub URL via the Agent Skills standard (agentskills.io).
  • PM Skills (opens in a new tab) - Product management marketplace with 48 agents and 65 skills spanning discovery, strategy, execution, analytics, GTM, and career development.
  • Planf3 (opens in a new tab) - Claude Code skill that generates structured HTML engineering plans from a reusable template, with embedded diagrams, validation checklists, and append-only amendment history; MIT.
  • Scrapling Agent Skill (opens in a new tab) - Official Agent Skills package bundling the Scrapling scraping library’s documentation as Markdown for Claude Code and OpenClaw, installable via Clawhub.
  • Skills for Antigravity (opens in a new tab) - Library of 462+ AI skills for Google Antigravity covering development, AI/ML (prompt engineering, RAG, agents), and business domains, ready to drop into a workspace.
  • Spec-Driven Orchestration Notes (opens in a new tab) - DivineDominion’s gist of spec-driven Claude Code prompts: requirements-discovery interview phases, developer and reviewer subagent definitions, and task-team orchestrator setup.
  • Supabase Agent Skills (opens in a new tab) - Supabase’s roughly 100-line agent skill for database, auth, and RLS work that teaches agents to fetch current docs instead of replicating them.
  • Taches CC Resources (opens in a new tab) - Personal collection of Claude Code commands and skills for planning, research methodology, MCP integration, and thinking frameworks.
  • Trail of Bits Skills (opens in a new tab) - Trail of Bits’ security skills for Claude Code: static analysis, smart-contract auditing, Semgrep rules, binary reverse engineering, and supply-chain checks.
  • Unblocked Skills (opens in a new tab) - Unblocked’s Agent Skills for unified context retrieval across code, pull requests, issues, docs, and team conversations, with semantic search; no license specified.
  • Vercel Agent Skills (opens in a new tab) - Vercel’s agent skills bundle: react-best-practices (40+ performance rules), web-design-guidelines (100+ a11y/UX rules), and vercel-deploy-claimable for instant framework-detected deploys.
  • nWave (opens in a new tab) - Claude Code agent framework that walks projects through six phases (requirements, architecture, testing, implementation) with mandatory human checkpoints and TDD enforcement.
  • Production subagent collection (opens in a new tab) - Collection of 80+ production-ready Claude Code subagents spanning languages, frameworks, and infrastructure, installable individually or via plugins.

Hooks & Automation

GUIs & Desktop Apps

  • AgentCraft (opens in a new tab) - RTS-game-style dashboard for orchestrating Claude Code agents; install via npx @idosal/agentcraft.
  • AionUi (opens in a new tab) - Free, local, open-source desktop app that auto-detects and unifies CLI agents (Claude Code, Codex, Hermes) into one multi-agent interface.
  • Claude Canvas (opens in a new tab) - TUI toolkit that gives Claude Code a second display for interactive interfaces (emails, calendars, flight bookings) using Bun and tmux split panes.
  • Claude Code Templates repository (opens in a new tab) - CLI tool for configuring and monitoring Claude Code with agents, commands, settings, hooks, and analytics.
  • Claudish (opens in a new tab) - BYOK coding assistant routing Claude Code to 15+ direct providers and 580+ models via OpenRouter, plus offline local models.
  • Clawdbot (opens in a new tab) - Self-hosted personal AI assistant connecting Claude and ChatGPT to WhatsApp, Telegram, Slack, and Discord; voice, browser automation, local-first.
  • NanoClaw (opens in a new tab) - Lightweight containerized alternative to OpenClaw on Anthropic’s Agents SDK; connects to WhatsApp, Telegram, Slack, Discord, Gmail with memory and scheduled jobs.
  • Open Design (opens in a new tab) - Open-source, BYOK alternative to Anthropic’s Claude Design that wires 16 detected agent CLIs into a local skill-driven design workflow producing HTML, PDF, and deck artifacts; Apache-2.0.
  • Supacode (opens in a new tab) - Native macOS terminal command center for coding agents and git worktrees, built in Swift on The Composable Architecture and libghostty; requires macOS 26+.
  • opcode (opens in a new tab) - Tauri desktop GUI for Claude Code with a visual session and project browser, sandboxed custom agents, MCP server management, usage and cost tracking, and checkpoint-based session time travel.

Status Lines & Trackers

  • Advanced tracking (opens in a new tab) - Real-time Claude Code usage monitor with predictions and warnings.
  • Agent Flow (opens in a new tab) - VS Code extension that visualizes Claude Code agent orchestration in real time, showing agents thinking, branching, and coordinating.
  • CC Status Line (opens in a new tab) - Statusline for Claude Code showing context usage, costs, and session metrics.
  • CCNotify (opens in a new tab) - macOS desktop notifications for Claude Code, alerting when input is needed or a task completes, with click-to-jump back to the session.
  • CLAudit (opens in a new tab) - Local CLI that aggregates Claude Code sessions across machines and projects, finds patterns, tracks file changes, and exports insights.
  • Claude Statusline (opens in a new tab) - Real-time Claude Code monitor with live cost calculation, 100+ powerline themes, multi-model support (Opus, Sonnet, Haiku), Git integration, and budget alerts.
  • CodexBar (opens in a new tab) - macOS menu-bar app that keeps AI coding-provider usage limits visible without logging in, tracking balances, spend, and quota-reset countdowns across 30+ services.
  • Irrlicht (opens in a new tab) - macOS menu-bar app showing one ambient state dot per AI coding session across Claude Code, Codex, Pi, Aider, and OpenCode; transcript-driven via FSEvents, no hooks.
  • React Grab (opens in a new tab) - Click any web element to copy filename, React component, and HTML source for Claude Code, Cursor, and Codex; up to 3x faster AI assistance.
  • Usage Monitoring (opens in a new tab) - CLI that analyzes Claude Code and Codex CLI usage from local JSONL files.
  • abtop (opens in a new tab) - htop/btop-style Rust TUI for Claude Code and Codex CLI sessions, tracking tokens, context window, rate limits, and ports.
  • codeburn (opens in a new tab) - TUI dashboard tracking AI coding token cost across Claude Code, Codex, and Cursor.
  • toktrack (opens in a new tab) - Rust token and cost tracker unifying Claude Code, Codex CLI, Gemini CLI, and OpenCode, with fast parsing, a persistent cache, and TUI views.

Routing & Proxies

  • Claude Code Proxy (opens in a new tab) - Proxy that runs Claude Code against OpenAI-compatible models, swapping the model backend without changing the agent harness.
  • Claude Code Router (opens in a new tab) - Routes Claude Code requests so you decide how to interact with the model while keeping Anthropic updates.
  • Shannon (opens in a new tab) - CLI and SDK wrapper from HumanLayer that drives a real interactive Claude Code session inside tmux, emitting stream-JSON for programmatic use.
  • Weave Router (opens in a new tab) - Transparent prompt router for Claude Code and Cursor that picks the most cost-effective model per request based on task complexity, claiming ~70% token-cost reduction.
  • ccundo (opens in a new tab) - Granular undo for Claude Code that reads session files to selectively revert file operations with preview and cascading safety.

Token & Context Optimization

  • Caveman (opens in a new tab) - Claude Code skill that cuts output tokens 65-87% by forcing terse caveman-style responses across four intensity levels, with hooks for 40+ AI coding agents.
  • CodeGraph (opens in a new tab) - Pre-indexed local code knowledge graph exposed via MCP, giving agents call paths, blast-radius analysis, and auto-synced context across 30+ languages.
  • Headroom (opens in a new tab) - Local-first context compression that cuts agent token use 60-95% on tool outputs, logs, and conversation history, with reversible compression, Apache 2.0.
  • Lynkr (opens in a new tab) - HTTP proxy for Claude Code, Cursor, and Codex that cuts token use up to 87.6% via output compression, semantic caching, and tier routing; Apache 2.0.
  • MCP Compressor (Atlassian Labs) (opens in a new tab) - Proxy MCP server that wraps existing MCP servers and replaces dozens of tool schemas with a small set of wrappers, cutting tokens by 70-97%.
  • RTK (opens in a new tab) - CLI proxy that cuts LLM token use 60-90% on common dev commands via pattern-based shell output compression, shipped as a single Rust binary.
  • Token Optimizer (opens in a new tab) - Tool that detects and removes wasted context (“ghost tokens”) in AI coding assistants, targeting verbose runtime output and structural waste; runs fully local, PolyForm Noncommercial license.
  • ctx (opens in a new tab) - Recommendation engine that watches development work and suggests relevant skills, agents, MCP servers, and harnesses from a 102k-node knowledge graph; MIT.
  • lean-ctx (opens in a new tab) - Single Rust binary that cuts LLM token use 60-95% via a shell hook plus MCP server, with AST-based signature extraction across 18 languages.

Project Management

  • Abacus (opens in a new tab) - Web UI for Beads-enabled repos, adding browser-based issue tracking, bulk updates, email notifications, and workflow automation on top of the CLI.
  • Beans (opens in a new tab) - CLI flat-file issue tracker where AI agents read, write, and resolve tickets inside the repo without an external service.
  • Claude Sessions (opens in a new tab) - Custom slash commands for Claude Code that track development sessions and document progress across them.
  • Claude Task Master (opens in a new tab) - AI task-management system that drops into Cursor, Lovable, Windsurf, Roo, and Claude Code to break work into tracked tasks.
  • FP (Fiberplane) (opens in a new tab) - Issue tracker built for coding agents, with scoped issues, durable context across sessions, and a CLI agents drive directly.
  • GuardKit (opens in a new tab) - Quality-gated planning tool with automatic task decomposition, SOLID/DRY/YAGNI checks, 80% test-coverage thresholds, TDD support, and Conductor.build integration.
  • Jira-to-PR (opens in a new tab) - CLI and MCP server that turns Jira tickets into GitHub pull requests via Claude, OpenAI, or Ollama, with preview, guardrails, and iterative refinement.
  • Modem (opens in a new tab) - Auto-triage tool that aggregates support, chat, and analytics signals into developer-ready issues with prioritization.
  • PM Brain (opens in a new tab) - Git-versioned PM operating system bundling frameworks, guides, templates, and playbooks for day-to-day product execution.
  • Plannotator (opens in a new tab) - Visual review tool for coding-agent plans and diffs, with team sharing and one-click feedback back to the agent.
  • Rezonant (opens in a new tab) - Product workspace that shapes ideas into engineering-ready tasks, syncing with Jira, Linear, and GitHub, with an isolated coding agent; closed source, paid.
  • Shotgun CLI (opens in a new tab) - Spec-driven CLI for technical research and spec generation aimed at engineers and AI code-gen tools.
  • Vibe Kanban (opens in a new tab) - Kanban-style runner for orchestrating multiple coding agents (Claude Code, Codex, others) with task tracking and parallel execution.

Learning & Courses

In this section

  • Authoring skills and subagents

    Our position on writing skills and subagents for Claude Code: selection first, evals before instructions, few maintained skills, and both guides.

  • Context and output interfaces

    Context pressure drops when each turn carries less and delegated work runs in its own window: load MCP servers per session, keep CLAUDE.md short and clear the window before it fogs. On the output side, HTML has become a practical default for specs and reports that people are meant to read.

  • Ecosystem and extension model

    Claude Code keeps its core harness thin, a single loop that leaves planning to the model, and moves specialization into plugins, MCP servers, SDKs and language servers. Native features keep absorbing popular extensions, so every extension choice needs regular review.

  • Evaluation and economics

    Claude Code's capability stories run ahead of production reality, and evals on your own historical tasks are the corrective. Token cost explains much of the tool market around it: trackers, routing proxies and compressors exist because agentic coding broke request-based pricing.