We love the Claude ecosystem.
So we built this.

ClaudeWave is a free, real-time directory of the Claude ecosystem (skills, subagents, MCP servers, hooks, templates, and tools). We index over 3,955 GitHub repositories and rank them by stars, velocity, and activity.

The problem: every week a dozen new MCP servers, skills, or agents pop up on GitHub. They are scattered across a hundred awesome-* lists and half of them are outdated. We kept losing an hour a day hunting for tools that already existed. So we built the directory we wished had existed.

What makes it different

• Broader than MCP-only directories. We cover 8 categories: skills, agents, MCP, hooks, plugins, templates, tools, and awesome-lists.
• Ranked by signal, not just stars. Velocity (stars/week), recency of commits, and heat score. Our Legendary tier rewards momentum, not just age.
• Use-case filters. Find tools by what you are building (creative, productivity, dev, data, marketing, automation…).
• Install snippets. Copy a ready-to-paste claude_desktop_config.json block for any MCP server.
• Stack Builder. Pick multiple servers, get one combined config file.

Built by ElephantPink

ClaudeWave is built and maintained by ElephantPink, an AI engineering team that ships custom Claude integrations, MCP servers and AI agent workflows for product companies across Europe, from gaming platforms to enterprise dashboards. We curate this directory from the same trenches we work in: real production Claude deployments, real MCP code, real agent skills shipped to clients.

What we build for clients

ElephantPink ships Claude work into production. If your team is exploring an integration but lacks Claude or MCP expertise in-house, this is what we cover end to end:

• Custom MCP servers. Bespoke Model Context Protocol servers exposing your data, APIs or tools to Claude. Stdio or HTTP transport, read-only or read-write, with permissions, auth and observability built in.
• Claude API integrations. Production integrations with Claude Code, Anthropic SDK or Managed Agents. Prompt caching, batch jobs, files, citations, tool use, structured outputs.
• Agent skills and subagents. Custom skills (.claude/skills/), subagents and slash commands tailored to your workflow. Versioning, distribution, permissions hardening.
• RAG and knowledge bases. Retrieval pipelines for support, sales or internal knowledge, with semantic search, evals and answer auditing.
• AI workflow automation. End-to-end automation of repetitive ops (content, lead qualification, ticket triage) using Claude plus your stack.

Need a custom Claude integration, MCP server or agent built for your product? Talk to ElephantPink

Frequently asked

How it works

Every day, a cron syncs the directory. We query the GitHub Search API with 19 curated queries (topics + names), parse 12 known awesome-list READMEs, and scrape GitHub Trending. Everything gets filtered (no forks, no archived, no spam), classified into 8 categories and 19 use cases, and scored. The data lives in a JSON file served directly from the build.

The front-end is Next.js 16 + Tailwind v4. No heavy backend: the sync pipeline runs on cron and the site is mostly static (except the snippet generator and search). That means it is fast, cheap, and open to contributions.

Legal bits

• ClaudeWave is not affiliated with, endorsed by, or sponsored by Anthropic PBC. "Claude" is a product name used descriptively under nominative fair use.
• All repository data is pulled from the public GitHub API. Star counts, descriptions, and READMEs belong to their original authors and are shown under the licenses published in each repo.
• If you are a maintainer and want your repo removed, email hello@claudewave.com. We will take it down, no questions asked.