Foggy Data MCP

Appearance

Foggy Data MCPStop Letting AI Write SQL Directly

Use a semantic layer and MCP tools to let AI query business data safely, with permissions, business meaning, and multi-database support built in.

中文快速开始
English Quick Start
GitHub
🔒

Safer Than Raw SQL

AI sends JSON DSL and MCP tool calls, not handwritten SQL. Keep dangerous operations and schema leakage out of the prompt.

🧠

Business Semantics First

TM/QM models expose dimensions, measures, permissions, and field meaning so AI works with business concepts instead of raw tables.

🤖

MCP-Native

Connect Claude Desktop, Cursor, Trae, and your own apps through MCP endpoints, metadata tools, query tools, and chart tools.

🌐

One Layer, Many Backends

Run the same query model across MySQL, PostgreSQL, SQL Server, SQLite, MongoDB, and language runtimes in Java or Python.

54+Built-in Functions
20+Query Operators
5Database Dialects
2Language Engines

What Problem Does It Solve?

Foggy Data MCP sits between AI and your database so the model asks business questions instead of generating fragile SQL.

Without a Semantic Layer

  • LLM prompts need table names, fields, and database dialect details
  • Permissions are hard to preserve once AI writes SQL directly
  • JOINs, aggregations, and business meanings drift quickly
  • Every new data source adds more prompt complexity

With Foggy Data MCP

  • AI calls MCP tools and sends JSON DSL instead of raw SQL
  • TM/QM models define business fields, relationships, and access rules
  • The engine handles JOINs, dialect translation, and query safety
  • You can ship one governed interface to multiple AI clients

Where To Start

Pick the path that matches your first real user and your fastest demo.

GENERAL MCP

通用数据语义层

适合要把 AI 安全接到业务库、报表库或分析库的团队。先跑通 MCP、DSL、图表和多数据库能力。

查看中文快速开始 →FASTEST GO-TO-MARKET

Odoo 智能问数场景

最容易拿到首批用户的切口。你已经有权限映射、模型和集成路径，适合直接做案例演示和行业推广。

查看 Odoo Bridge →

How It Works

Natural language or app requests are turned into governed semantic queries before touching the database.

text
AI Assistant / App
        │
        │  MCP tools / JSON-RPC
        ▼
Foggy MCP Server
  • metadata tool
  • query tool
  • chart tool
        │
        ▼
Semantic Layer
  • TM/QM models
  • JSON Query DSL
  • permission injection
  • dialect SQL generation
        │
        ▼
MySQL / PostgreSQL / SQL Server / SQLite / MongoDB

Three Reasons People Try It

SECURITY

Keep AI Away From Raw SQL

  • Read-only query flow
  • Permission injection before execution
  • No need to dump your schema into prompts
  • Cleaner audit and governance boundaries
SEMANTICS

Expose Business Meaning

  • Dimensions, measures, hierarchies, calculated fields
  • Reusable TM/QM modeling
  • One query language across data sources
  • Less prompt engineering, more stable results
ADOPTION

Meet Users Where They Already Are

  • Claude Desktop, Cursor, Trae, custom apps
  • Java and Python implementations
  • Chart generation for demos and reporting
  • Good fit for ERP, BI, and internal AI tools

Sample Use Cases

Lead with a scenario, not with the framework. These are easier to market than a generic platform pitch.

📊

AI 数据分析助手

让业务人员直接问“上周各品牌销售趋势”“本月退款率最高的品类”，避免他们碰 SQL。

🏢

Odoo 智能问数

把 Odoo 权限规则映射到 DSL 查询，做销售、采购、库存、财务的自然语言分析。

🛡️

受控企业 AI 接口

给内部 Copilot、客服机器人或 BI 助手一层可治理的数据访问协议，而不是直接数据库连接。

Choose Language / 选择语言

The docs are bilingual. Pick your working language and keep moving.

🇨🇳简体中文中文文档与示例🇺🇸EnglishEnglish docs and references
☕ Java - Spring Boot 🐍 Python - FastAPI 🧩 Odoo - ERP Scenario

Start With One Demo, Not a Huge Platform Rollout

The fastest path is to publish one visible scenario, one short video, and one reproducible quick start. Then iterate from real user questions.

中文文档English DocsGitHub