Teradata MCP 服务

Teradata MCP Server 是一款将 AI 代理直接安全连接到 Teradata 数据库的 MCP 服务,支持自然语言查询与分析、语义层构建、RAG 管道及数据库管理,适用于 AI 助手进行数据探索、分析和运维的场景。

  • by Teradata
  • Python
  • 本地部署
  • 58 星标
  • 更新于
在 GitHub 查看 Teradata/teradata-mcp-server

使用说明

Teradata MCP Server

Connect AI agents directly to Teradata with enterprise security and extensibility.

Quick Start (Choose Your Path)

Client Best For Setup Time
Claude Desktop Exploratory analysis, platform admin 5 min
VS Code + Copilot Data engineering, agent development 5 min
Open WebUI Testing new LLMs locally 5 min
Code Examples Build your own client varies
Flowise Visual agent builder 10 min

Pre-requisites: Teradata database (or free sandbox) + uv

Claude Desktop Setup (No Installation)

Add this to claude_desktop_config.json (Settings > Developer > Edit Config):

{
  "mcpServers": {
    "teradata": {
      "command": "uvx",
      "args": ["teradata-mcp-server"],
      "env": {
        "DATABASE_URI": "teradata://<USERNAME>:<PASSWORD>@<HOST_URL>:1025/<USERNAME>"
      }
    }
  }
}

What You Can Do

Use Case Capabilities Tools
Query & Analyze Explore tables, profile data, explain results, visualize patterns—no SQL needed base, dba, qlty, plot
Semantic Layer Generate custom semantic layers and tools from YAML or with our Agent Skill Learn more →
AI & RAG Pipelines Semantic search, retrieval-augmented generation, vector storage rag, tdvs, fs
Database Admin Manage security, monitor capacity, automate backups dba, sec, bar

What's New (Latest Release)

  • FastMCP v4 — Sessionless protocol, background tasks for long-running analytics, argument completion
  • Response Caching — 5-minute TTL signals reduce redundant database queries in multi-turn conversations
  • Guard Mode — Multi-step confirmation flows for destructive operations (bar_, sec_)
  • Background Tasks — tdml_* analytic functions return task IDs for polling instead of blocking
  • Argument Completion — Auto-suggest table and column names from schema as users type
  • Hooks Capability — Intercept tool calls for custom monitoring, audit, or rate-limiting
  • Row Limit Protection — Configurable caps (DEFAULT_ROW_LIMIT, MAX_ROW_LIMIT) prevent LLM token overflow
  • Enhanced Security — VX views for fine-grained row-level access control

Extend & Deploy

Add Custom Logic
Use hooks to intercept tool calls for monitoring, audit trails, or validation → Hooks Guide

Define Semantic Layers
Create domain-specific tools, prompts, and cubes in YAML → Customization Guide

Deploy Everywhere
Run as CLI (uv), HTTP server, Docker container, or cloud service → Installation Guide

See It In Action

Learn More

Contributing

We welcome contributions! See our Contributing Guide and Developer Guide to get started.