Skip to main content

MCP Tools

Available MCP tools for storing, searching, managing memories, and working with audio/video content.

Overview

The Trix MCP server exposes several tools that AI assistants can use to manage memories. These tools are automatically available to any MCP-compatible client once configured.

You don't need to call these tools directly. AI assistants automatically select the appropriate tool based on your natural language requests.

Tool Categories

  • Memory Tools - Store, recall, list, and manage memories
  • Audio Tools - Transcribe, search, and analyze audio/video content
  • Topic Tools - Work with LLM-extracted topics
  • Cluster Tools - Organize memories into semantic groups
  • Enrichment Tools - Add LLM-generated metadata to memories

store_memory

Store a new memory in Trix. Use this tool to save important information for later recall.

Parameters

ParameterTypeRequiredDescription
contentstringYesThe content to store as a memory
tagsstring[]NoTags to categorize the memory
metadataobjectNoAdditional metadata to attach
session_idstringNoLink memory to a specific session for tracking conversation context

Example

store_memory
// Tool: store_memory
// Parameters:
{
  "content": "The user prefers dark mode in all applications",
  "tags": ["preference", "ui"],
  "metadata": {
    "source": "conversation",
    "confidence": 0.95
  }
}

// Returns:
{
  "id": "mem_abc123",
  "content": "The user prefers dark mode in all applications",
  "created_at": "2024-01-15T10:30:00Z"
}

Example Prompts

  • "Remember that I prefer TypeScript over JavaScript"
  • "Save this: my favorite color is blue"
  • "Store this information for later: meeting with John at 3pm tomorrow"

search_memories

Search for relevant memories using semantic similarity. This tool finds memories that are conceptually related to your query, even if they don't contain the exact words.

Parameters

ParameterTypeRequiredDescription
querystringYesThe search query
limitnumberNo (default: 10)Maximum number of results
thresholdnumberNo (default: 0.5)Minimum similarity score (0-1)

Example

search_memories
// Tool: search_memories
// Parameters:
{
  "query": "What are the user's UI preferences?",
  "limit": 5,
  "threshold": 0.7
}

// Returns:
{
  "memories": [
    {
      "id": "mem_abc123",
      "content": "The user prefers dark mode in all applications",
      "similarity": 0.92,
      "tags": ["preference", "ui"]
    }
  ]
}

There is no dedicated list tool. Use search_memories with a broad query to browse stored memories, or use list_pinned_memories / list_protected_memories for filtered listing.

delete_memory

Delete a specific memory. Use this when information is no longer relevant or was stored incorrectly.

Parameters

ParameterTypeRequiredDescription
memory_idstringYesID of the memory to delete

Example

delete_memory
// Tool: delete_memory
// Parameters:
{
  "id": "mem_abc123"
}

// Returns:
{
  "success": true,
  "deleted_id": "mem_abc123"
}

Memory Management Tools

Tools for managing memory lifecycle, including pinning, protection, and soft deletion.

pin_memory

Pin a memory to prevent it from decaying over time. Pinned memories maintain their priority and are always included in relevant search results.

ParameterTypeRequiredDescription
memory_idstringYesThe ID of the memory

unpin_memory

Unpin a previously pinned memory, allowing it to decay naturally.

protect_memory

Set the protection level for a memory to prevent accidental deletion.

ParameterTypeRequiredDescription
memory_idstringYesThe ID of the memory to protect
level"none" | "soft" | "hard"YesProtection level: none removes protection, soft warns before deletion, hard prevents deletion

soft_delete_memory

Soft delete a memory, allowing it to be restored later. Soft-deleted memories are hidden from normal queries but can be recovered.

restore_memory

Restore a soft-deleted memory, making it visible and searchable again.

Soft-deleted memories are automatically purged after 30 days. Use restore_memory to recover them before the purge period ends.

Topic Tools

Tools for working with LLM-extracted topics from memories.

get_memory_topics

Get the topics extracted from a memory. Topics are automatically categorized and scored for relevance.

ParameterTypeRequiredDescription
memory_idstringYesThe ID of the memory
refreshbooleanNo (default: false)Force re-extraction of topics
min_relevancenumberNoMinimum relevance score (0.0-1.0)
categoriesstring[]NoFilter by categories: technology, business, science, people, places, events, concepts

search_by_topic

Find memories that are related to a specific topic.

ParameterTypeRequiredDescription
topicstringYesThe topic to search for
limitnumberNo (default: 10)Maximum number of results

Cluster Tools

Tools for working with automatically generated memory clusters.

list_clusters

List clusters at a specific scale. Clusters group related memories together.

ParameterTypeRequiredDescription
scale"fine" | "medium" | "coarse"No (default: medium)Cluster scale
limitnumberNo (default: 20)Maximum number of clusters to return

get_cluster_details

Get details about a specific cluster, optionally including its member memories.

ParameterTypeRequiredDescription
cluster_idstringYesThe ID of the cluster
include_memoriesbooleanNo (default: false)Include member memories in the response

Cluster scales determine granularity: fine creates small, specific clusters (3-10 memories), medium creates balanced clusters (10-50 memories), and coarse creates large thematic clusters (50+ memories).

Clustering Tools

Tools for triggering cluster generation to organize memories.

trigger_clustering

Trigger cluster generation for memories. This organizes memories into semantic groups at different scales (fine, medium, coarse) for better discovery.

ParameterTypeRequiredDescription
space_idstringNoLimit clustering to a specific space
memory_idsstring[]NoSpecific memory IDs to include in clustering

Enrichment Tools

Tools for enriching memories with LLM-generated metadata.

enrich_memory

Enrich a memory with LLM-generated metadata including topics, summaries, entities, and quality scores.

ParameterTypeRequiredDescription
memory_idstringYesID of the memory to enrich
operationsstring[]NoEnrichment types: "topics", "summary", "entities", "quality", "facts"

Audio Tools

Tools for working with audio and video content, including transcription, search, and segment extraction.

Audio tools work with uploaded audio and video files. Transcription is performed automatically or can be triggered manually. All transcripts are indexed for semantic search.

list_audio_files

List audio files with optional filters.

ParameterTypeRequiredDescription
limitnumberNo (default: 20)Maximum number of results
offsetnumberNo (default: 0)Number of results to skip for pagination
space_idstringNoFilter by space ID
transcription_status"pending" | "processing" | "completed" | "failed"NoFilter by transcription status

get_audio_transcript

Get the full transcript for an audio file. Optionally include detailed segment information and speaker identification.

ParameterTypeRequiredDescription
audio_file_idstringYesThe ID of the audio file
include_segmentsbooleanNo (default: false)Include timestamped segments
include_speakersbooleanNo (default: false)Include speaker diarization info

search_audio_transcripts

Search within audio transcripts using semantic search. Returns matches with timestamps, speaker information, and clip URLs for precise navigation.

ParameterTypeRequiredDescription
querystringYesThe search query
limitnumberNo (default: 10)Maximum number of results
thresholdnumberNo (default: 0.5)Minimum similarity score (0-1)
space_idstringNoFilter by space ID

get_audio_segments

Get transcript segments with timestamps and speaker information.

ParameterTypeRequiredDescription
audio_file_idstringYesThe ID of the audio file
speakerstringNoFilter by speaker ID (e.g., "speaker_1")
start_timenumberNoStart time in seconds
end_timenumberNoEnd time in seconds

trigger_transcription

Start a transcription job for an audio or video file.

ParameterTypeRequiredDescription
memory_idstringYesThe memory ID of the uploaded audio/video file
languagestringNoLanguage code (e.g., "en", "es", "fr"). Auto-detected if not specified.
enable_diarizationbooleanNo (default: true)Enable speaker diarization to identify different speakers
enable_entity_detectionbooleanNo (default: false)Detect and tag entities like names, dates, locations

Transcription jobs run asynchronously. Processing time depends on audio duration and selected options. Diarization and entity detection add additional processing time but provide richer metadata.

Updated Tool Parameters

Several existing tools have been enhanced with new parameters for pinning, protection, quality filtering, and cluster awareness.

store_memory (Updated)

The store_memory tool now accepts additional parameters for memory management:

ParameterTypeDescription
is_pinnedbooleanPin the memory immediately upon creation
protection_level"none" | "soft" | "hard"Set initial protection level

search_memories (Updated)

The search_memories tool now supports additional filtering options:

ParameterTypeDescription
cluster_scale"fine" | "medium" | "coarse"Filter results by cluster scale
min_qualitynumberMinimum quality score (0.0-1.0)
pinnedbooleanFilter to only pinned memories
protectedbooleanFilter to only protected memories

list_memories (Updated)

The list_memories tool now supports additional filtering options:

ParameterTypeDescription
pinnedbooleanFilter to only pinned memories
protectedbooleanFilter to only protected memories
min_qualitynumberMinimum quality score (0.0-1.0)
include_deletedbooleanInclude soft-deleted memories in results

Next Steps

  • Advanced - Troubleshooting and advanced configuration
  • Daemon - Run Trix as a background service