MCP Server Features

The TigerGraph MCP server exposes a rich set of tools to AI applications, allowing them to "operate" the database through natural language.

Core Functionalities

The official TigerGraph and community-led (Custom Discoveries) MCP implementations provide the following tools:

1. Schema Introspection

Allows the LLM to retrieve the full graph schema, including vertex types, edge types, and their respective attributes. This is the first step for an AI to "understand" what questions it can ask.

2. Query Management

• Query View: Lists all available installed GSQL queries that the AI is authorized to run.
• Query Execution: The AI can execute installed queries or raw GSQL strings with parameters passed from a natural language prompt.

3. Data Manipulation

• Create Entities: Programmatically create vertices and edges.
• Upsert Attributes: Update existing attributes, including support for Vector attributes (crucial for hybrid search).
• Alter Schema: Dynamically alter vertices to add support for new vector dimensions.

4. Advanced Tools

• UDF Listing: Fetch installed User-Defined Functions (UDFs).
• Metric Retrieval: Advanced implementations allow AI to query system performance, such as CPU and memory usage, directly via natural language (e.g., "Run CPU & memory usage").

[!TIP] Use Schema Introspection at the beginning of a session to give the LLM the context it needs to construct valid GSQL parameters.