Advanced Usage

Beyond simple queries, MCP allows for complex AI orchestration through agent frameworks.

MCP ChatBot

Both the official and community versions include a CLI-based ChatBot. This allows you to "chat" with your database directly from the terminal.

bashterminal
Copy
# Example interaction
> "Find the shortest path between User_1 and User_10"
> "Show me the distribution of vertex types"

AG2 Agent Framework (v3.1+)

The TigerGraph MCP version 3.1 (by Custom Discoveries) is built on the AG2 agent framework (formerly AutoGen).

• Native LLM Support: AG2 allows the server to support multiple LLM families (OpenAI, Anthropic, etc.) without complex conditional code.
• Tool Discovery: The system can automatically identify which MCP server (if multiple are connected) has the specific tool needed to fulfill a request.

Integration with CrewAI

For multi-agent workflows, TigerGraph can be used as a "Researcher" or "Data Analyst" agent within CrewAI.

• Example: One agent fetches graph metrics while another agent uses those metrics to generate a business report.

[!TIP] Check the Roadmap for upcoming features like streaming support and enhanced multi-graph security within MCP.