io.github.deficlow由 GitHub 账号 deficlow 发布v0.1.1更新于 Oct 24, 2026
Search and inspect 6,500+ curated AI apps from the HyperStore directory.
发布方自己的服务：Zipp
把 HyperStore 收录的 6500 多款 AI 应用目录开放给助手，可按关键词、意图、分类、人群或用例搜索、筛选和比较工具。
Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。
把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。
阅读 https://sourceweft.com/mcp/io.github.deficlow%2Fhyperstore-mcp/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。把它添加到你客户端的 mcpServers 配置中。
{
"mcpServers": {
"hyperstore": {
"type": "http",
"url": "https://mcp.store.hypergpt.ai/mcp"
}
}
}Plug 6,500+ AI apps into any LLM via the Model Context Protocol.
[PyPI] [Glama] [Smithery] [MCP Registry] [CI] [License: MIT]
HyperStore is a curated directory of 6,500+ AI applications, developed by HyperGPT. This MCP server exposes the HyperStore catalog to any LLM client — Claude, ChatGPT, Cursor, Windsurf, Cline, Zed, Gemini, and anything else that speaks MCP.
Ask your LLM:
"Find me a free AI tool that summarises PDFs." "Compare ChatGPT, Claude, and Gemini side-by-side." "Show me the top 5 image-generation apps with an API."
The LLM calls HyperStore MCP behind the scenes and answers with up-to-date, curated results.
13 tools:
3 resources:
hyperstore://app/{slug} — markdown rendering of any apphyperstore://category/{slug} — top apps in a categoryhyperstore://catalog — full category index3 prompts:
find_tool_for_task — guided discovery for a taskcompare_apps — side-by-side app comparisondiscover_category — explore a topicuvx (zero install, recommended)Requires uv. One command and you're done:
pipxUse our managed Streamable HTTP server:
Edit ~/Library/Application Support/Claude/claude_desktop_config.json
(macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
Restart Claude → tools appear in the 🛠 menu.
.cursor/mcp.json (project) or ~/.cursor/mcp.json (global):
~/.codeium/windsurf/mcp_config.json:
settings.json:
~/.config/zed/settings.json:
~/.gemini/settings.json:
Settings → Connectors → Add custom connector:
https://mcp.store.hypergpt.ai/mcpSee examples/ for ready-to-paste configs for every supported client.
For self-hosting, use the Docker image.
For direct invocation without Docker, the CLI accepts --transport http|sse
(see hyperstore-mcp --help).
When self-hosting, these environment variables can be set
(see .env.example for the full list):
Inspect the running server with the official MCP Inspector:
HyperStore MCP is a thin async wrapper around the HyperStore public REST API. It is read-only — no credentials, no writes, no PII. The same data that powers the website powers the MCP server. Updates land in your LLM the moment they land on the site.
MIT © HyperGPT
来源：README.md，提交 0b66aad
Search HyperStore's AI apps directory by keyword. Returns a paginated list of matching apps with name, slug, short description, pricing, and rating. Use this when the user gives concrete keywords (e.g. 'image upscaler', 'code copilot').
limitquery必填cursorNatural-language semantic search powered by embeddings. Best for fuzzy intent ('a tool that helps me write Python tests', 'something like Midjourney but free'). Returns up to 12 apps ranked by semantic similarity.
prompt必填Fetch the full detail page for a single AI app by slug: long description, features, screenshots, categories, pricing, rating, website URL, source attribution.
slug必填Paginated apps listing with optional filters. Combine `category`, `pricing`, and a free-text `query` to drill down. Returns apps sorted by popularity. Use `cursor` (last app id from previous page) to paginate.
limitquerycursorpricingcategoryList all HyperStore categories with app counts. Use this first when the user asks 'what kinds of AI tools are there?' or to discover available category slugs.
无需输入参数。
Get apps within a specific category. Returns the category metadata plus a paginated list of apps in that category, sorted by popularity.
slug必填limitcursorpricingBrowse apps A-Z by starting letter. Use letter='#' for apps starting with digits or symbols. Useful for alphabetical discovery rather than search.
limitcursorletter必填pricingFetch the HyperStore homepage payload: top categories with their featured apps, the trending apps strip, and totals. Good first call to give the user a broad overview.
pageGet curated alternatives to a specific AI app by slug. Returns the app plus a list of competing/similar tools ranked by match confidence. Use this when the user asks 'what are alternatives to X', 'something like X', or 'X vs others'.
slug必填limitList the audience segments HyperStore curates tools for (e.g. 'developers', 'lawyers', 'students'), each with a slug and app count. Call this first to discover audience slugs for `apps_for_audience`.
无需输入参数。
Get the best AI tools for a specific audience by slug (from `list_audiences`), ranked by relevance then popularity. Use when the user asks 'best AI tools for {role/profession}'. Paginate with `cursor` (last app id from the previous page).
slug必填limitcursorList the use-case taxonomies HyperStore curates tools for (e.g. 'legal-contracts', 'tiktok-shorts'), each with a slug and app count. Call this first to discover use-case slugs for `apps_for_use_case`.
无需输入参数。
Get AI tools for a specific use case by slug (from `list_use_cases`), ranked by relevance then popularity. Use when the user asks 'AI tools for {task}'. Paginate with `cursor` (last app id from the previous page).
slug必填limitcursor已满足适用检查项的 100%
该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。
uvx hyperstore-mcppipx install hyperstore-mcphyperstore-mcpdocker run --rm -p 8080:8080 ghcr.io/deficlow/hyperstore-mcp# Now MCP Streamable HTTP at http://localhost:8080/mcphttps://mcp.store.hypergpt.ai/mcp{ "mcpServers": { "hyperstore": { "command": "uvx", "args": ["hyperstore-mcp"] } }}claude mcp add hyperstore -- uvx hyperstore-mcp{ "mcpServers": { "hyperstore": { "command": "uvx", "args": ["hyperstore-mcp"] } }}{ "mcpServers": { "hyperstore": { "command": "uvx", "args": ["hyperstore-mcp"] } }}{ "cline.mcpServers": { "hyperstore": { "command": "uvx", "args": ["hyperstore-mcp"] } }}{ "context_servers": { "hyperstore": { "command": { "path": "uvx", "args": ["hyperstore-mcp"] } } }}{ "mcpServers": { "hyperstore": { "command": "uvx", "args": ["hyperstore-mcp"] } }}from openai import OpenAI
client = OpenAI()response = client.responses.create( model="gpt-4.1", tools=[{ "type": "mcp", "server_label": "hyperstore", "server_url": "https://mcp.store.hypergpt.ai/mcp", "require_approval": "never", }], input="Find me 3 free AI tools for writing unit tests.",)print(response.output_text)from anthropic import Anthropic
client = Anthropic()response = client.messages.create( model="claude-opus-4-7", max_tokens=1024, mcp_servers=[{ "type": "url", "url": "https://mcp.store.hypergpt.ai/mcp", "name": "hyperstore", }], messages=[{"role": "user", "content": "Top 5 AI image generators?"}],)git clone https://github.com/deficlow/HyperStore-MCPcd HyperStore-MCPuv sync --all-extrasuv run pytestuv run hyperstore-mcp # stdio mode for local testingnpx @modelcontextprotocol/inspector uvx hyperstore-mcpLLM client ──MCP──▶ hyperstore-mcp ──HTTPS──▶ store.hypergpt.ai/api