Teradata MCP Server is an MCP (Model Context Protocol) server developed by the Teradata community, with the goal of connecting AI agents directly to Teradata databases while maintaining scalability alongside enterprise-grade security. The project provides a set of tools that agents can invoke, centered on data quality management and read/write operations on Teradata databases.
Key capabilities
- FastMCP v4: Uses a sessionless protocol, supports background tasks for long-running analytical tasks, and supports parameter completion.
- Response caching: Through a 5-minute TTL signal, reduces repeated database queries in multi-turn conversations.
- Guard Mode: Provides a multi-step confirmation process for destructive operations (prefixed with bar_, sec_).
- Background tasks: tdml_* analysis functions return a task ID for polling instead of blocking and waiting.
- Parameter completion: Automatically suggests table names and column names as the user types, based on the database schema.
- Hooks: Can intercept tool calls for custom monitoring, auditing, or rate limiting.
- Row limit protection: Configurable capacity limits (DEFAULT_ROW_LIMIT, MAX_ROW_LIMIT) to prevent LLM token overflow.
- Enhanced security: Implements fine-grained row-level access control through VX views.
Usage
Prerequisites: a Teradata database (or free sandbox) and uv.
Claude Desktop setup (no installation required): Add the corresponding configuration to claude_desktop_config.json (Settings > Developer > Edit Config).
Extension and deployment
- Add custom logic: Use hooks to intercept tool calls for monitoring, audit trails, or validation.
- Define semantic layers: Create domain-specific tools, prompts, and cubes in YAML.
- Deploy anywhere: Can run as a CLI (uv), HTTP server, Docker container, or cloud service.
Examples and documentation
The project showcases examples such as voice agents (real-time bidirectional audio using Amazon Nova Sonic), web agents (interactive chat UI using the Google ADK framework), Flowise builder (visual drag-and-drop workflow), custom middleware (performance monitoring patterns), and provides complete documentation (installation, configuration, architecture,security), video tutorials, developer guides, and architecture descriptions. The project welcomes contributions and provides contribution guidelines and a developer guide.
Applicable scenarios and limitations
This server is suitable for scenarios that require AI agents to securely access Teradata databases, perform data quality management, and carry out read/write operations. Its enterprise-grade security features (Guard Mode, VX views, row limits) help control risk. Specific installation steps, configuration details, and complete capabilities require reference to the official documentation; specific operation commands, pricing, and platform support information not provided in this introduction are not yet available.
