Trueai

io.github.chenyujing1234-netizenPublished by GitHub account chenyujing1234-netizenv1.0.1Updated Oct 24, 2026

Recommend and inspect AI SaaS apps from the TrueAI catalog (1,600+ apps) by need, name or URL.

VerifiedStreamable HTTPWeb executableWeb Search & ScrapingAI & ML

Overview

AI-generated overview

A remote MCP server that lets an assistant recommend and look up AI SaaS apps from the TrueAI catalog of 1,600+ tools by need, name, URL or category.

What you can do
Recommend AI tools from a natural-language description of a need, scenario, audience or budget, with a configurable number of results and optional review snippets attached to a tool's record. Look up one app's full structured record by display name, slug, numeric id or official URL. Browse the catalog with filters such as category, free tier and form factor, plus sorting. List every catalog category with its slug, name and tool count for use as a filter.
When to use it
Useful when someone wants help choosing an AI tool or SaaS app and would rather have the assistant search a curated catalog than the open web, or when you need structured details such as pricing, ratings and links for a specific AI product.
Requirements
A remote streamable HTTP endpoint; no package install, account, API key or header is declared. Network access to the provider's host is required.
Before you install
The catalog is third-party curated data, so recommendations and ratings should be treated as opinions rather than verified facts. The README describes automated collection of app data by crawling sites with a logged-in browser, so listing details may be incomplete or out of date. The tools are read-only lookups; no writes or payments are exposed.

Installation

In SourceWeft

  1. Open Trueai in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/io.github.chenyujing1234-netizen%2Ftrueai/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "trueai": {
      "type": "http",
      "url": "https://www.shiflowai.cloud/mcp"
    }
  }
}

README

真选AI (TrueAI) — 新一代 AI 导航，懂你

[Live] [MCP] [Skill] [StructAIWeb] [License: MIT]

我来帮你省钱，我来帮你省时间。 数据实时、人工评测、无广告。 没有最好的，只有最适合你的。

TrueAI 是一个面向 AI 工具的 智能体推荐 + 人工评测 + 对话式导航 平台 MVP。 用户可以：

  • 浏览按类目组织的 AI 工具卡片墙（带吊绳轻摆动效）
  • 用对话的方式让 "懂你" 助手反问并推荐（流式 SSE）
  • 查看多维排行榜，支持"用自然语言生成专属排名"
  • 把最多 3 个工具加入对比抽屉，做并排多维对比
  • 阅读 / 提交工具评测（审核通过后可获得现金奖励；MVP 仅保留表结构）

🤖 接入你的 AI Agent（MCP / Skill）

TrueAI 1600+ 应用目录已封装成 MCP Server 和 单文件 Skill，让你的 agent 直接获得"挑 AI 工具"的能力。完整文档：skills/trueai/SKILL.md。

一行接入 MCP（Claude Desktop / Cursor / Cline / Continue / Windsurf）

json
{ "mcpServers": { "trueai": { "url": "https://www.shiflowai.cloud/mcp" } } }

接入后 agent 立即获得 4 个工具：

工具用途
recommend_ai_tools(description, top_k)用户自然语言需求 → 推荐合适应用
get_ai_tool(name_or_url, include_reviews)名称 / slug / 官网 URL → 应用完整信息
list_ai_tools(category, free_only, ...)按分类 / 形态 / 是否免费 浏览目录
list_categories()列出全部分类

无 MCP 也能用——drop-in skill 文件

把 skills/trueai/SKILL.md 复制到任意 agent 的 skills 目录， agent 就会按文档里 curl 示例直接调 https://www.shiflowai.cloud/api/*。

数据 schema

每个 AI 应用 34 个结构化字段（评分 / 价格 / 形态 / 评论 / 链接 …）， 完整 JSON Schema：docs/ai_tool_schema.json。

🕷️ 数据从哪里来：StructAIWeb（核心采集智能体）

TrueAI 网站里 1600+ 应用的结构化字段——名称、Logo、Slogan、定价、子功能列表、 易用性 / 效果 / 性价比评分等——不是人工录入的，而是由我们另一个开源项目 StructAIWeb 自动跑出来的。 它是 TrueAI 数据管线的 核心 AI 智能体：

维度说明
仓库https://github.com/chenyujing1234-netizen/StructAIWeb
形态Windows 桌面 GUI + CLI（Python + Playwright）
浏览器通过 CDP 接管你本地已登录的 Chrome，天然复用 Cookie / 扩展，规避大多数风控
探索启发式按「价格 → 功能 → 关于 → 文档」遍历站点，自动点 CTA 进入主功能区
试用主动输入提示词、点 "生成 / 提交"，永远避开 "购买 / 充值 / 订阅" 按钮
视觉可选启用多模态视觉模型（Qwen-VL / GPT-4o / GLM-4V）识别"该点哪个按钮"
登录墙自动检测并暂停，等你完成登录后继续
整合把所有页面快照交给 LLM 输出结构化 JSON + 三项主观评分（usability / effect / price）
输出字段与 docs/ai_tool_schema.json 完全对齐
LicenseMIT

简言之：

StructAIWeb 负责"看懂一个 AI 网站"，TrueAI 负责"把上千个看懂的结果织成一个懂你的导航站"。

想自己跑数据 / 扩展更多 AI 工具？fork StructAIWeb，跑完把 output/*.json 喂进 TrueAI 的 backend/scripts/sync_*.py 入库脚本即可。

技术栈

层选型
后端Python 3.10+ / FastAPI 0.115 / SQLAlchemy 2 / Alembic / PyMySQL
前端Next.js 14 (App Router, 纯 JavaScript) / Tailwind CSS / Framer Motion / Zustand
数据库MySQL 8 (114.55.254.123:3306/trueai)
LLM阿里云 DashScope (qwen-plus)，未配置 Key 时自动回退 Mock 文案
鉴权JWT (python-jose) + bcrypt

目录结构

TrueAI/├── backend/              # FastAPI 服务│   ├── app/│   │   ├── core/         # 配置 / DB / 安全│   │   ├── models/       # ORM 模型│   │   ├── schemas/      # Pydantic│   │   ├── api/routers/  # auth / categories / tools / reviews / rankings / stats / search│   │   ├── services/llm/ # Qwen 与 Mock 两种 Provider│   │   ├── services/recommend.py  # 意图抽取 + MySQL 候选召回│   │   └── seeds/seed_tools.py│   ├── alembic/│   └── requirements.txt└── frontend/             # Next.js 14    ├── app/              # layout / page / category / tool / search / rankings / compare / earn    ├── components/       # NavBar / HeroTypewriter / Sidebar / ToolCard / CompareBar / ChatSearch ...    ├── lib/              # api.js / compareStore.js / labels.js    └── tailwind.config.js

快速开始

前提：已安装 Python 3.10+、Node 20+、能访问 MySQL 114.55.254.123:3306。

1. 后端

bash
cd backendpython3 -m venv venvsource venv/bin/activatepip install -r requirements.txtcp .env.example .env    # 按需改 JWT_SECRET / DASHSCOPE_API_KEY
# 首次：执行数据库迁移 + 种子数据alembic upgrade headpython -m app.seeds.seed_tools
uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload# http://localhost:8000/docs 查看 OpenAPI

2. 前端

bash
cd frontendnpm installcp .env.local.example .env.local   # 默认指向 http://127.0.0.1:8000npm run dev# 打开 http://localhost:3000

环境变量

backend/.env

变量说明
DATABASE_URL形如 mysql+pymysql://user:pass@host:3306/trueai，密码中特殊字符需 URL 编码（@ → %40）
JWT_SECRETJWT 签名密钥，生产请替换为长随机串
JWT_EXPIRES_MINUTEStoken 过期分钟数，默认 7 天
LLM_PROVIDERqwen（默认）。为空或 Key 缺失时自动回退 Mock
DASHSCOPE_API_KEY通义千问 Key，在 DashScope 控制台 创建
QWEN_MODEL默认 qwen-plus，可换成 qwen-turbo / qwen-max
CORS_ORIGINS允许的前端 origin，逗号分隔

frontend/.env.local

变量说明
NEXT_PUBLIC_API_BASE后端 Base URL，开发默认 http://127.0.0.1:8000

前端所有 /api/* 请求在 Next.js 端通过 rewrites 代理到后端，避免浏览器 CORS 问题。

核心 API

接口说明
GET /api/health健康检查
GET /api/stats首页横幅用：工具数 / 分类数 / 评测数 / 价值观文案
GET /api/categories全部分类 + 每类工具数
GET /api/tools列表 + 多维筛选 + 排序 + 分页
GET /api/tools/{id 或 slug}智能体详情
POST /api/tools/compare对比（body 为 id 列表）
GET /api/rankings?dimension=&category=&top=排行榜
POST /api/reviews提交评测（需 Bearer token）
GET /api/reviews?tool_id=&status_filter=approved评测列表
POST /api/search/chat对话搜索，SSE 流：event: meta + 多个 event: delta + event: done
GET /api/search/chat?q=...非流式兜底：直接给候选和外部导航
POST /api/auth/register / /login / GET /api/auth/me注册 / 登录 / 查询当前用户

种子数据自带一个管理员账号：admin / admin123456（生产请删除或改密）。

"懂你"对话搜索的工作方式

  1. 客户端把当前会话（含历史）POST 到 /api/search/chat。
  2. 后端根据最新 user 消息，用关键词字典抽取 类目 / 人群 / 免费 / 国内直连 等意图。
  3. 先从 MySQL 召回 ≤ 6 个候选，立即通过 event: meta 推送给前端（前端马上渲染卡片）。
  4. 接着把系统 Prompt + 候选摘要 + 会话历史交给 Qwen 流式输出，按字/片推送 event: delta。
  5. 若候选为空，meta 中带 external 外部 AI 导航站清单，前端展示"去别家看看"。

常见问题

  • ValueError: invalid interpolation syntax：Alembic 读 % 会报错。alembic/env.py 已做 % → %% 转义。
  • pymysql 连接慢 / 超时：pool_pre_ping=True、pool_recycle=3600 已打开；请确认 MySQL 白名单已允许你的出口 IP。
  • 前端 API 跨域：开发期通过 next.config.js rewrites 代理，无需额外处理。部署时建议把后端放到同域 /api/* 路径下。
  • Qwen 没配 Key：/api/search/chat 会自动回退到 Mock Provider，候选卡片依然是真实数据，只是 AI 文案是占位。

下一步（非 MVP）

  • 任务 / 赏金系统：管理员发布任务 → 用户领取评测 → 后台审核 → 微信提现
  • 用户自发布任务
  • 评测通过后自动刷新工具的综合评分
  • 微信登录 + 手机号登录
  • 工具增删改的管理后台

MIT License.

Source: README.md at commit 24cb234

Tools

4

recommend_ai_tools

Find AI tools that match a natural-language user need. Parameters ---------- description : str The user's need in their own words. Be specific about the scenario, audience, platform, language and budget when known. Examples: '我是新手想学编程，要免费的 AI 工具'; 'I need an AI to generate a 10-page PPT from an outline'. top_k : int (1-10, default 5) Maximum number of candidates to return. Returns ------- JSON: {intents, candidates, external}. ``candidates`` is already ranked by relevance & score. ``external`` lists fallback nav sites only when no in-catalog candidate matches.

Parameterstop_kdescriptionrequired

get_ai_tool

Resolve and return the full structured record of one AI app. Parameters ---------- name_or_url : str The AI app's display name (e.g. 'Cursor', '通义千问'), its TrueAI slug, a numeric id, or its official URL (e.g. 'https://cursor.com'). include_reviews : bool (default false) If true, attach up to ``max_reviews`` real user reviews (sourced from watcha.cn). Useful for social proof / real-world usage tips. max_reviews : int (1-20, default 5) Cap on attached reviews. Resolution order: numeric id → exact slug → official_url substring → brand keyword from URL → fuzzy name/tagline match.

Parametersmax_reviewsname_or_urlrequiredinclude_reviews

list_ai_tools

Browse the catalog with structured filters and sorting. Parameters ---------- category : str | None Optional category slug to filter by, e.g. 'coding', 'image-gen', 'writing'. Use list_categories() to discover slugs. free_only : bool | None If true, only return tools with a free tier. form_factor : str | None Filter by product form: 'saas' | 'mobile' | 'cli' | 'windows_app' | 'web' | 'browser_extension' | 'api'. sort : str (default 'overall') Sort key: 'overall' | 'usability' | 'effect' | 'price' | 'speed' | 'reviews' | 'new'. page : int (>=1, default 1) page_size : int (1-100, default 20) Returns ``{items, total, page, page_size}``. Each item is a brief record; call ``get_ai_tool`` to drill in.

Parameterspagesortcategoryfree_onlypage_sizeform_factor

list_categories

List every category in the TrueAI catalog. Each entry has ``slug``, ``name``, optional ``icon``, ``sort_order``, ``parent_id`` and ``tool_count``. Use the returned slugs to filter ``list_ai_tools(category=...)``.

Takes no parameters.

Version history

1
  1. v1.0.1LatestSep 16, 2026

Score

100% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listedRequired15 pts
  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts