Skip to main content

Available Tools

The Tensorheart Memory MCP Server provides Claude with 9 powerful tools for managing memories. Here's what Claude can do:

Memory Operations​

Creating Memories

Memories are created through the ingest_document tool or the extraction API. This ensures high-quality memories with automatic entity extraction, deduplication, and conflict resolution.

See Advanced Operations for document ingestion, or use the API extraction endpoint for text content.

query_memories​

Search memories using natural language.

Usage:

Claude, what are my preferences for application settings?

Parameters:

  • query (required) - Natural language query
  • space (optional) - Limit search to specific space
  • max_memories (optional) - Maximum memories to return (1-50, default: 10)
  • threshold (optional) - Relevance threshold 0.0-1.0 (default: 0.5)

What happens:

  • Searches your memories for relevant information
  • Returns the most relevant results
  • Only memories above the relevance threshold are returned

update_memory​

Update an existing memory.

Usage:

Claude, update memory mem_abc123 to say I prefer light mode instead.

Parameters:

  • memory_id (required) - ID of the memory to update
  • text (optional) - New text content
  • space (optional) - New space for the memory

What happens:

  • Memory content or location is updated
  • Updated timestamp is set
  • Memory remains searchable with new content

delete_memory​

Delete a memory permanently.

Usage:

Claude, delete memory mem_abc123.

Parameters:

  • memory_id (required) - ID of the memory to delete

What happens:

  • Memory is permanently deleted
  • Cannot be recovered
  • Removed from all searches

get_memory​

Retrieve a specific memory by ID.

Usage:

Claude, show me memory mem_abc123.

Parameters:

  • memory_id (required) - ID of the memory to retrieve

What happens:

  • Full memory details are returned
  • Includes text, entities, timestamps, and metadata

Space Management​

create_space​

Create a new memory space for organizing memories.

Usage:

Claude, create a memory space called "work-projects".

Parameters:

  • name (required) - Space name (alphanumeric, hyphens, underscores, spaces)
  • description (optional) - Space description

What happens:

  • New space is created
  • Can be used to organize future memories
  • Spaces can be queried independently

Example spaces:

  • personal - Personal notes and preferences
  • work - Work-related information
  • project-alpha - Project-specific context
  • learning - Things you're learning

list_spaces​

List all your memory spaces.

Usage:

Claude, what memory spaces do I have?

What happens:

  • Returns all spaces with their names, descriptions, and memory counts
  • Includes the default space
  • Shows which spaces have the most memories

get_space​

Get details about a specific space.

Usage:

Claude, show me details about the "work-projects" space.

Parameters:

  • space_id (required) - ID of the space to retrieve

What happens:

  • Returns space details including memory count
  • Shows when the space was created
  • Indicates if it's the default space

Advanced Operations​

consolidate_memories​

Find and merge similar memories to reduce clutter.

Usage:

Claude, consolidate my memories in the "work" space.

Parameters:

  • space (optional) - Limit consolidation to specific space
  • threshold (optional) - Merge sensitivity 0.0-1.0 (default: 0.8)
  • dry_run (optional) - Preview changes without applying (default: true)

What happens:

  • Finds duplicate or overlapping memories
  • Shows which memories would be merged
  • If dry_run=false, actually merges them

Best practices:

  • Always run with dry_run=true first to preview
  • Higher threshold (0.9) = only merge very closely related memories
  • Lower threshold (0.7) = merge more broadly related memories

ingest_document​

Process a document and create memories from it.

Usage:

Claude, ingest the file at /path/to/document.pdf into my "documentation" space.

Parameters:

  • file_path (required) - Path to the document file
  • space (optional) - Space to store memories
  • title (optional) - Document title

Supported formats:

  • PDF documents
  • Images (JPEG, PNG)
  • Audio files (MP3, WAV)
  • Text files

What happens:

  • Document is uploaded and processed
  • Content is extracted from the file
  • Creates searchable memories from the content

Example use cases:

  • Upload API documentation for reference
  • Process meeting transcripts
  • Make research papers searchable
  • Convert lecture notes into memories

Resources​

In addition to tools, the MCP server provides read-only resources that Claude can access:

memory://all​

List all memories with pagination

memory://recent​

List recently created/updated memories

memory://spaces​

List all memory spaces

memory://spaces/{space_id}​

List memories in a specific space

memory://entities​

List all extracted entities

memory://entities/{type}​

List entities of a specific type (PERSON, ORG, LOCATION, SKILL, PRODUCT)

Tool Combinations​

Common Workflows​

1. Process and Search Documentation:

1. Claude, create a space called "api-docs"
2. Claude, ingest this API documentation PDF into "api-docs" space
3. Claude, how do I authenticate with this API?
4. Claude, what rate limits does this API have?

2. Build a Knowledge Base from Documents:

1. Claude, create a space called "team-info"
2. Claude, ingest our team directory PDF into "team-info"
3. Claude, who is our project manager?
4. Claude, what's Bob's role?

3. Process Meeting Notes:

1. Claude, ingest this meeting transcript into "meetings" space
2. Claude, what action items were discussed?
3. Claude, what decisions were made about the timeline?

4. Clean Up Duplicates:

1. Claude, consolidate memories in the "work" space (dry run)
2. [Review the suggestions]
3. Claude, consolidate memories in the "work" space without dry run

Tips for Effective Memory Usage​

Use Document Ingestion​

✅ Ingest documents, transcripts, and notes for best results
✅ Let the extraction system identify and structure important information

Use Spaces​

✅ Create separate spaces for different projects or topics
✅ Query within specific spaces for more relevant results

Regular Consolidation​

✅ Run consolidation monthly to remove duplicates
✅ Always use dry_run first to preview changes

Leverage Entities​

✅ Ingested content automatically extracts entities
✅ Query by person, organization, or skill names

Next Steps​