Skip to main content

Examples and Use Cases

Real-world examples showing how to use Tensorheart Memory MCP with Claude Desktop.

Personal Knowledge Management​

Building Your Second Brain​

Track everything you learn and let Claude help you recall it:

Storing Knowledge:

You: I just learned that the Rust ownership system uses move semantics by default,
which means values are moved rather than copied unless the type implements Copy trait.

Claude: I'll remember that for you.
[Uses create_memory tool]

Done! I've saved this information about Rust's ownership system to your memory.

Recalling Knowledge Later:

You: What did I learn about Rust ownership?

Claude: Let me check your memories...
[Uses query_memories with query="Rust ownership"]

Based on your notes, Rust's ownership system uses move semantics by default,
meaning values are moved rather than copied unless the type implements the Copy trait.

Meeting Notes Organization​

Create spaces for different projects and store meeting notes:

You: Create a space called "project-alpha" for my new project

Claude: I'll create that space for you.
[Uses create_space with name="project-alpha"]

Done! Created space "project-alpha" for organizing your project memories.

---

You: We decided in today's meeting that we'll use PostgreSQL for the database
and Redis for caching. Store this in project-alpha.

Claude: I'll save that decision to your project-alpha space.
[Uses create_memory with space="project-alpha"]

Saved! Your architectural decisions are now stored in the project-alpha space.

Code Development Assistance​

Tracking Project Patterns​

Remember coding patterns and standards across conversations:

You: In this project, we always use dependency injection for services
and store configs in a config/ directory. Remember this.

Claude: I'll remember these project conventions.
[Uses create_memory]

Saved! I'll keep these patterns in mind when helping with this project.

---
[Next conversation]

You: I need to add a new email service

Claude: Let me check your project patterns...
[Uses query_memories with query="service patterns dependency injection"]

Based on your project conventions, I'll implement the email service using
dependency injection. Should I create the config in the config/ directory?

API Documentation Memory​

Store API endpoints and usage examples:

You: Our API uses /api/v1/users for user management.
POST creates a user, GET lists them, GET /:id retrieves one.
All requests need Authorization: Bearer {token} header.

Claude: I'll save this API documentation.
[Uses create_memory]

Saved! I now have your API endpoints and auth requirements in memory.

---
[Later]

You: Show me how to create a user in our API

Claude: Let me recall the API docs...
[Uses query_memories with query="API users endpoint POST"]

Based on your API documentation, here's how to create a user:

```bash
curl -X POST https://api.example.com/api/v1/users \
  -H "Authorization: Bearer YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"name": "John Doe", "email": "john@example.com"}'

## Research and Learning

### Document Ingestion for Study Materials

Ingest PDFs and create searchable knowledge:

You: I have a research paper on machine learning. Can you ingest this PDF at ~/Downloads/ml-research-2024.pdf into my "research" space?

Claude: I'll process that document for you. [Uses create_space if needed, then ingest_document]

Done! I've processed the paper and created searchable memories from it in your research space. You can now ask me questions about the content.

You: What does the paper say about transformer architectures?

Claude: Let me search your research memories... [Uses query_memories with query="transformer architectures" and space="research"]

According to the paper you uploaded, transformer architectures use self-attention mechanisms to...


### Entity Tracking

Automatically track people, organizations, and locations:

You: I met Sarah Chen from DataCorp at the AI conference in San Francisco. She mentioned they're using GPT-4 for customer service.

Claude: I'll save this information. [Uses create_memory - automatically extracts entities]

Saved! I've stored this with entities:

  • Person: Sarah Chen
  • Organization: DataCorp
  • Location: San Francisco
  • Product: GPT-4

[Later]

You: Who did I meet from DataCorp?

Claude: Let me check... [Uses query_memories with query="DataCorp person"]

You met Sarah Chen from DataCorp at the AI conference in San Francisco. She mentioned they're using GPT-4 for customer service.


## Content Creation

### Writer's Notes and Ideas

Track story ideas, character details, and plot points:

You: For my sci-fi novel: The protagonist, Dr. Elena Martinez, discovers that quantum entanglement can transmit consciousness. She works at CERN. Save this to my "novel" space.

Claude: I'll save this character and plot detail. [Uses create_memory with space="novel"]

Saved to your novel space! I've noted Dr. Elena Martinez and her quantum consciousness discovery.

[Next session]

You: What was my protagonist's name and where does she work?

Claude: Let me check your novel notes... [Uses query_memories with query="protagonist character" and space="novel"]

Your protagonist is Dr. Elena Martinez, and she works at CERN where she discovers that quantum entanglement can transmit consciousness.


## Memory Consolidation

### Cleaning Up Duplicate Information

Find and merge overlapping memories:

You: I think I've saved information about Python typing multiple times. Can you consolidate related memories?

Claude: I'll check for overlapping memories and show you what can be merged. [Uses consolidate_memories with dry_run=true]

I found these overlapping memory groups:

  1. "Python uses type hints with typing module" and "Python typing module for type annotations"
  2. "Use mypy for type checking" and "mypy is the Python type checker"

Should I merge these? This will combine them into clearer, consolidated memories.

You: Yes, merge them

Claude: I'll merge the related memories now. [Uses consolidate_memories with dry_run=false]

Done! Merged 2 groups of overlapping memories. Your memory store is now cleaner and more organized.


## Workspace Organization

### Multi-Space Workflow

Organize different areas of your life:

You: Create spaces for my work projects, personal learning, and home projects

Claude: I'll create those spaces for you. [Uses create_space three times]

Created 3 spaces:

  • work-projects
  • personal-learning
  • home-projects

You can now organize memories by area of your life.

You: List my spaces

Claude: Let me show you your spaces. [Uses list_spaces]

You have 3 memory spaces:

  1. work-projects (5 memories)
  2. personal-learning (12 memories)
  3. home-projects (3 memories)

You: Save "Need to fix the leaky faucet in bathroom" to home-projects

Claude: I'll save that task. [Uses create_memory with space="home-projects"]

Saved to home-projects! Your home task list now includes fixing the bathroom faucet.


## Advanced Workflows

### Cross-Space Research

Query across all spaces for comprehensive recall:

You: Search all my memories for anything about "authentication"

Claude: I'll search across all your spaces. [Uses query_memories without specifying space]

Found authentication-related memories across multiple spaces:

From work-projects:

  • "Use JWT tokens with 15-minute expiry and refresh tokens"
  • "OAuth 2.0 flow: authorization code grant for web apps"

From personal-learning:

  • "Learned about bcrypt for password hashing with cost factor 12"
  • "Session-based auth vs token-based auth trade-offs"

### Knowledge Export Pattern

Get comprehensive memory dumps for backup or review:

You: Show me all my recent memories

Claude: I'll retrieve your recent memories. [Uses list_recent_memories resource via memory://recent]

Here are your 20 most recent memories: [Lists all recent memories with timestamps]

You can review what you've been learning and working on recently.


## Tips for Effective Use

1. **Be Specific**: Include context when storing memories
   - Good: "In React 18, use useTransition hook for non-urgent updates to keep UI responsive"
   - Less helpful: "React has hooks"

2. **Use Spaces**: Organize by project, topic, or area
   - Easier to find relevant information later
   - Prevents mixing work and personal knowledge

3. **Regular Consolidation**: Run consolidation monthly
   - Keeps memories clean and organized
   - Merges duplicate or overlapping information

4. **Ingest Documents**: Upload PDFs, notes, and references
   - Creates searchable knowledge base
   - No need to manually copy/paste content

5. **Trust the Search**: Use natural language queries
   - "What did I learn about databases?" works better than single keywords
   - Ask questions naturally - the system finds what you need

6. **Update When Needed**: Don't create new memories for updates
   - Use update_memory to revise existing information
   - Keeps knowledge current without duplication