Skip to main content

MCP Server for Claude

The Tensorheart Memory MCP Server brings persistent memory to Claude Desktop and Claude Code. Instead of forgetting everything between conversations, Claude can now remember, recall, and build upon your interactions over time.

What is MCP?​

MCP (Model Context Protocol) is Anthropic's standardized protocol for connecting Claude to external tools and data sources. The Tensorheart Memory MCP Server implements this protocol to give Claude access to your Tensorheart Memory account.

What Can Claude Do with Memory?​

Once configured, Claude can:

  • 📝 Remember - Save important information, preferences, and context
  • 🔍 Recall - Search through memories using natural language
  • 🗂️ Organize - Group memories into spaces (like folders)
  • 🧹 Consolidate - Merge similar memories to reduce clutter
  • 📄 Ingest - Turn documents (PDFs, images, audio) into searchable memories

Real-World Examples​

Personal Assistant​

"Claude, remember that I prefer tabs over spaces in Python code." "Claude, what are my coding preferences?"

Project Memory​

"Claude, create a space called 'project-alpha'" "Claude, remember that we decided to use PostgreSQL for this project" "Claude, what database are we using for project-alpha?"

Document Processing​

"Claude, ingest this API documentation PDF" "Claude, how do I authenticate with the XYZ API?"

How It Works​

┌─────────────────┐
│ Claude Desktop  │  You interact with Claude normally
│  / Claude Code  │
└────────┬────────┘
         │
         │ MCP Protocol
         ▼
┌──────────────────────┐
│ Tensorheart Memory   │  Runs on your computer
│    MCP Server        │
└──────────┬───────────┘
           │
           │ HTTPS
           ▼
┌──────────────────────┐
│ Tensorheart Memory   │  Your memories stored securely
│        API           │
└──────────────────────┘
  1. MCP Server runs locally on your machine
  2. Claude communicates with the MCP server using MCP protocol
  3. MCP server connects to Tensorheart Memory API using your API key
  4. Your memories are stored securely in the cloud

Key Features​

Claude can search your memories using natural language queries - just ask what you want to find.

Memory Spaces​

Organize memories into namespaces (like "work", "personal", "project-alpha") to keep things organized.

Entity Extraction​

Automatically identifies people, organizations, locations, skills, and products mentioned in your memories.

Document Ingestion​

Upload PDFs, images, or audio files and Claude can search through them later.

Memory Consolidation​

Find and merge similar or duplicate memories to keep your memory store clean.

Getting Started​

Ready to give Claude a memory? Follow our Installation Guide to get started in just a few minutes.

Requirements​

  • Python 3.11 or higher
  • Claude Desktop or Claude Code
  • Tensorheart Memory API account (Sign up)

Next Steps​

  1. Install the MCP Server
  2. Configure Claude Desktop
  3. Explore Available Tools
  4. See Examples

Need Help?​