claude-oracle-mcp
An MCP server for discovering Claude Code skills, plugins, and MCP servers by searching 15,000+ resources from 17 registries, GitHub, and the web with zero setup.
README
<img align="right" src="claude-oracle.svg" alt="claude-oracle-mcp" width="220">
claude-oracle-mcp
An Model Context Protocol (MCP) server for discovering Claude Code skills, plugins, and MCP servers. Search 15,000+ resources from 17 registries + GitHub search + web search with zero setup.
<br clear="right">

install
Requirements:
From shell:
claude mcp add claude-oracle-mcp -- npx claude-oracle-mcp
From inside Claude (restart required):
Add this to our global mcp config: npx claude-oracle-mcp
Install this mcp: https://github.com/Vvkmnn/claude-oracle-mcp
From any manually configurable mcp.json: (Cursor, Windsurf, etc.)
{
"mcpServers": {
"claude-oracle-mcp": {
"command": "npx",
"args": ["claude-oracle-mcp"],
"env": {
"SKILLSMP_API_KEY": "optional-for-semantic-search",
"GITHUB_TOKEN": "optional-for-higher-rate-limits"
}
}
}
}
There is no npm install required -- no external dependencies or local databases, only search algorithms.
However, if npx resolves the wrong package, you can force resolution with:
npm install -g claude-oracle-mcp
skill
Optionally, install the skill to teach Claude when to proactively use oracle:
npx skills add Vvkmnn/claude-oracle-mcp --skill claude-oracle --global
# Optional: add --yes to skip interactive prompt and install to all agents
This makes Claude automatically search for relevant tools before planning, when encountering errors, or at session start. The MCP works without the skill, but the skill improves discoverability.
plugin
For automated tool discovery with hooks, install from the claude-emporium marketplace:
/plugin marketplace add Vvkmnn/claude-emporium
/plugin install claude-oracle@claude-emporium
The claude-oracle plugin provides:
Hooks (targeted, fires before external searches):
- Before WebSearch/WebFetch → Check oracle for existing tools first
- Before EnterPlanMode → Search for relevant plugins/skills
Command: /oracle-search <query>
Requires the MCP server installed first. See the emporium for other Claude Code plugins and MCPs.
features
MCP server that gives Claude access to 15,000+ skills, plugins, and MCP servers from 17 registries + GitHub search + web search. Fast discovery with smart prioritization.
Runs locally (with crystal vision 🔮):
search
Search across all resources for relevant tools and solutions.
🔮 search query=<query>
> "Are there any MCP servers for PostgreSQL?"
> "What testing frameworks exist for React?"
> "Do we have authentication plugins?"
🔮 search | postgres
╭─ 🔮 ─────────────────────────────── Found 5 ─╮
│ • postgres-mcp (mcp) │
│ PostgreSQL database access │
│ smithery.ai • ⭐ verified │
│ Install: npx postgres-mcp │
│ │
│ • postgresql-mcp (mcp) │
│ PostgreSQL MCP server │
│ npmjs.com • ✨ quality │
│ Install: npx @scope/postgresql-mcp │
│ │
│ Total: 19 sources • 15,000+ resources │
╰──────────────────────────────────────────────╯
{
"results": [
{
"name": "postgres-mcp",
"type": "mcp",
"description": "PostgreSQL database access",
"source": "smithery.ai",
"install_command": "npx postgres-mcp",
"verified": true,
"stars": 142
},
{
"name": "postgresql-mcp",
"type": "mcp",
"description": "PostgreSQL MCP server",
"source": "npmjs.com",
"install_command": "npx @scope/postgresql-mcp",
"quality_score": 0.85
}
],
"sources_searched": ["smithery.ai", "npmjs.com", "glama.ai", "modelcontextprotocol.io"],
"total_available": 15000,
"cached": false
}
browse
Browse resources by category, type, or popularity.
🔮 browse category=<category>
> "Browse testing tools"
> "Show MCP servers only"
> "Find security-related skills"
🔮 browse | testing
╭─ 🔮 ─────────────────────────────── Found 8 ─╮
│ • tdd-workflows (plugin) │
│ Test-driven development workflow │
│ claude-code-plugins-plus • 258 plugins │
│ │
│ • pytest-mcp (mcp) │
│ Python testing framework │
│ smithery.ai • ⭐ verified │
│ │
│ Total: 19 sources • 15,000+ resources │
╰──────────────────────────────────────────────╯
{
"results": [
{
"name": "tdd-workflows",
"type": "plugin",
"description": "Test-driven development workflow",
"source": "claude-code-plugins-plus",
"install_command": "git clone https://github.com/user/tdd-workflows ~/.claude/plugins/tdd-workflows",
"category": "testing"
},
{
"name": "pytest-mcp",
"type": "mcp",
"description": "Python testing framework",
"source": "smithery.ai",
"install_command": "npx pytest-mcp",
"verified": true
}
],
"sources_searched": ["claude-code-plugins-plus", "smithery.ai", "npmjs.com"],
"total_available": 15000,
"cached": true
}
sources
Show all available data sources and their status.
🔮 sources
> "Which sources are available?"
> "Check data source health"
🔮 sources | 19 total
╭─ 🔮 ─────────────────────────────── 15,000+ total ─╮
│ Plugins (303): │
│ • claude-code-plugins-plus: 258 ✓ │
│ • claude-plugins-official: 45 ✓ │
│ │
│ MCP Servers (14,358): │
│ • smithery.ai: 500 ✓ │
│ • playbooks.com: 1,000+ ✓ │
│ • npmjs.com: 250 ✓ │
│ • modelcontextprotocol.io: 248 ✓ │
│ • glama.ai: 662 ✓ │
│ • wong2/awesome-mcp-servers: 200+ ✓ │
│ • punkpeye/awesome-mcp-servers: 400+ ✓ │
│ • collabnix/awesome-mcp-lists: 100+ ✓ │
│ │
│ Skills (339): │
│ • awesome-agent-skills: 339 ✓ │
│ • awesome-claude-code: 200+ ✓ │
│ • jmanhype/awesome-claude-code: 50+ ✓ │
│ • skillsmp: 25,000 (optional key) │
╰────────────────────────────────────────────────────╯
{
"sources": [
{ "name": "claude-code-plugins-plus", "type": "plugin", "count": 258, "status": "ok" },
{ "name": "claude-plugins-official", "type": "plugin", "count": 45, "status": "ok" },
{ "name": "smithery.ai", "type": "mcp", "count": 500, "status": "ok" },
{ "name": "playbooks.com", "type": "mcp", "count": 1000, "status": "ok" },
{ "name": "npmjs.com", "type": "mcp", "count": 250, "status": "ok" },
{ "name": "modelcontextprotocol.io", "type": "mcp", "count": 248, "status": "ok" },
{ "name": "glama.ai", "type": "mcp", "count": 662, "status": "ok" },
{ "name": "awesome-agent-skills", "type": "skill", "count": 339, "status": "ok" },
{ "name": "skillsmp", "type": "skill", "count": 25000, "status": "no_key" }
],
"total": 15000
}
methodology
How claude-oracle-mcp works:
User: "What MCP servers exist for PostgreSQL?"
↓
┌─────────────────────┐
│ 🔮 claude-oracle │
│ MCP Server │
└─────────┬───────────┘
│
┌──────────────┼──────────────┐
│ PARALLEL SEARCH (19) │
└──────────────┬───────────────┘
│
┌───────────────────┼───────────────────┐
│ │ │
┌───▼────┐ ┌────▼─────┐ ┌────▼─────┐
│Smithery│ │ npm │ │ Glama.ai │
│ 500 │ │ 250 │ │ 662 │
└────────┘ └──────────┘ └──────────┘
│ │ │
└───────────────────┼───────────────────┘
│
┌─────────▼───────────┐
│ DEDUPLICATE & │
│ RANK RESULTS │
└─────────┬───────────┘
│
┌─────────▼───────────┐
│ FORMAT & RETURN │
│ 🔮 Beautiful │
│ Bordered Output │
└─────────────────────┘
Core features:
- 17 registries + GitHub search + web search: Smithery, npm, Glama.ai, GitHub, awesome lists + more
- 15,000+ resources: Skills, plugins, and MCP servers in one search
- Parallel fetching: All sources searched simultaneously (~3 seconds)
- Smart caching: In-memory TTL cache (6-24 hour expiry)
- Keyword scoring: Weighted matching (exact > starts with > contains)
- Quality signals: Stars, verified badges, quality scores boost ranking
- Beautiful formatting: Bordered output with 🔮 identifier
Data sources (zero-config, 16/17 registries):
| Source | Type | Count | Method |
|---|---|---|---|
| Smithery Registry | MCP | 500 | REST API (5 pages) |
| Playbooks | MCP | 1,000+ | XML Sitemaps |
| Official MCP Registry | MCP | 248 | REST API (3 iterations) |
| npm Registry | MCP/Plugin | 250 | Search API (1 page) |
| Glama.ai | MCP | 662 | RSS Feed |
| wong2/awesome-mcp-servers | MCP | 200+ | GitHub Markdown |
| punkpeye/awesome-mcp-servers | MCP | 400+ | GitHub Markdown |
| collabnix/awesome-mcp-lists | MCP | 100+ | GitHub Markdown |
| jmanhype/awesome-claude-code | Plugin/Skill | 50+ | GitHub Markdown |
| hesreallyhim/awesome-claude-code | Skill | 200+ | GitHub Markdown |
| awesome-agent-skills | Skill | 339 | GitHub Markdown |
| claude-code-plugins-plus | Plugin | 258 | Marketplace JSON |
| claude-plugins-official | Plugin | 45 | Marketplace JSON |
| superpowers-marketplace | Plugin | - | Marketplace JSON |
Query-based (searched per query, zero-config):
| Source | Type | Count | Method |
|---|---|---|---|
| GitHub (search) | MCP/Plugin | ~20 | REST API (per query) |
| Web (search) | MCP/Plugin | ~10 | DuckDuckGo Lite (per query) |
Optional (requires API key):
- SkillsMP: 25,000+ skills with semantic search (get key)
- GITHUB_TOKEN: Higher GitHub API rate limits (10 → 30 req/min)
Note: Pagination is limited for faster responses (~3s). Full datasets available through caching on subsequent searches.
Design principles:
- Aggregated search -- single query searches 17 registries + GitHub + web simultaneously
- Zero-config -- 18/19 sources work without API keys (GITHUB_TOKEN optional for rate limits)
- Deduplication -- cross-source result merging by name similarity
- Relevance scoring -- weighted name match, description match, and popularity
- Offline fallback -- graceful degradation when sources are unreachable
alternatives
Every MCP discovery tool either searches a single registry or requires separate accounts per source. Oracle aggregates 19 sources in one command.
| Feature | oracle | 1mcpserver | Single registry |
|---|---|---|---|
| Resource types | MCP + plugins + skills | MCP servers only | Usually one type |
| Sources | 19 (registries + awesome lists + GitHub + web) | Registry searches | One source at a time |
| Skills/plugins | Yes (339 skills, 303 plugins) | No | Usually no |
| Setup | One MCP, one command | One MCP, one command | Separate account/API per registry |
| Cross-source | Deduplicated, ranked results | Per-registry | Manual comparison |
| Caching | In-memory TTL (6-24h) | Per-request | Varies |
| Install commands | Included in results | Included in results | Varies |
1mcpserver -- MCP server discovery from multiple registries. Searches MCP servers only -- no plugins, no skills. Per-registry results without cross-source dedup or ranking. No caching between requests.
Smithery, Glama, SkillsMP -- Individual registries, each searchable independently. Requires separate accounts or API access per registry. Each covers a subset of the ecosystem -- no single source has everything. Manual comparison across registries to find the best option. Oracle aggregates all of these (and 16 more sources) with deduplicated, ranked results in one query.
development
git clone https://github.com/Vvkmnn/claude-oracle-mcp && cd claude-oracle-mcp
npm install && npm run build
npm test
Package requirements:
- Node.js: >=20.0.0 (ES modules)
- Runtime:
@modelcontextprotocol/sdk,fast-xml-parser - Zero external databases -- works with
npx
Development workflow:
npm run build # TypeScript compilation with executable permissions
npm run dev # Watch mode with tsc --watch
npm run start # Run the MCP server directly
npm run lint # ESLint code quality checks
npm run lint:fix # Auto-fix linting issues
npm run format # Prettier formatting (src/)
npm run format:check # Check formatting without changes
npm run typecheck # TypeScript validation without emit
npm run test # Lint + type check
npm run prepublishOnly # Pre-publish validation (build + lint + format:check)
Git hooks (via Husky):
- pre-commit: Auto-formats staged
.tsfiles with Prettier and ESLint
Contributing:
- Fork the repository and create feature branches
- Test with multiple data sources before submitting PRs
- Follow TypeScript strict mode and MCP protocol standards
Learn from examples:
- Official MCP servers for reference implementations
- TypeScript SDK for best practices
- Creating Node.js modules for npm package development
license
<hr>
<p align="center"> <a href="https://en.wikipedia.org/wiki/Cumaean_Sibyl"><img src="logo/oracle.jpg" alt="Aeneas and the Cumaean Sibyl -- Claude Mellan" width="340"></a> </p>
<p align="center">
Aeneas and the Cumaean Sibyl by Claude Mellan. The oracle who guided Aeneas through the underworld.
</p>
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。