---

license: mit title: Tensorus MCP sdk: python emoji: 🐠 colorFrom: blue colorTo: yellow short_description: Model Context Protocol server and client for Tensorus tensor database

Tensorus MCP

Model Context Protocol (MCP) server and client for Tensorus tensor database operations. This package provides a standardized interface for AI agents and LLMs to interact with Tensorus capabilities using the Model Context Protocol.

Features

• MCP Server: Python implementation using fastmcp for tensor database operations
• MCP Client: Python client library for easy integration with MCP servers
• Tensor Operations: Complete set of tensor manipulation tools via MCP
• Dataset Management: Create, list, and manage tensor datasets
• Demo Mode: Pre-configured mock data for testing and demonstration

Installation

pip install fastmcp
pip install -r requirements.txt

Quick Start

Starting the MCP Server

python -m tensorus_mcp.server

For web endpoint support:

python -m tensorus_mcp.server --transport streamable-http

Demo Mode

For demonstration or testing purposes, run the server in demo mode:

python -m tensorus_mcp.server --demo-mode

Using the Python Client

from tensorus_mcp.client import TensorusMCPClient

async def example():
    async with TensorusMCPClient.from_http("http://localhost:8000/mcp/") as client:
        # List available datasets
        datasets = await client.list_datasets()
        print(f"Available datasets: {datasets}")
        
        # Create a new dataset
        await client.create_dataset("my_dataset")
        
        # Ingest a tensor
        result = await client.ingest_tensor(
            dataset_name="my_dataset",
            tensor_shape=[2, 2],
            tensor_dtype="float32",
            tensor_data=[[1.0, 2.0], [3.0, 4.0]],
            metadata={"source": "example"}
        )
        print(f"Ingested tensor with ID: {result['record_id']}")

MCP Demo Script

Prerequisites

• Tensorus MCP Server running (python -m tensorus_mcp.server)
• For live mode: Tensorus backend API accessible
• For demo mode: No additional setup required

Demo Scenario: MCP Client Interaction

Goal: Demonstrate how an external AI agent can leverage Tensorus via MCP.

1. Start MCP Server:

   python -m tensorus_mcp.server --demo-mode
   

2. Connect via Python Client:

   from tensorus_mcp.client import TensorusMCPClient
   
   async def demo():
       async with TensorusMCPClient.from_http("http://localhost:8000/mcp/") as client:
           # List available datasets
           datasets = await client.list_datasets()
           print(f"Available datasets: {datasets}")
           
           # Create a new dataset
           result = await client.create_dataset("demo_dataset")
           print(f"Created dataset: {result}")
           
           # Ingest sample tensor data