memvid-mcp

Generate an audit report with sources and citations. Creates a detailed report showing which sources support a query. Useful for fact-checking and citation generation. Args: file: Path to the .mv2 memory file query: Query topic for the audit top_k: Number of sources to include include_snip…

Debug segment information (internal structure). Shows internal index structure for debugging. Segment types: - lex: Lexical/full-text index (Tantivy) - vec: Vector index (embeddings) - time: Temporal index Args: file: Path to the .mv2 memory file segment_type: Type of segment to debug Return…

Export memory data to JSON, CSV, or JSONL format. Args: file: Path to the .mv2 memory file output: Output file path format: Export format (json, csv, jsonl) frame_ids: Export only specific frame IDs (optional) Returns: Export status and output file path

List internal SQLite tables and structures. Shows database schema for advanced debugging. Args: file: Path to the .mv2 memory file Returns: List of tables with column definitions

Schema operations - infer or show schema summary. Can infer schemas from stored data or show existing schema definitions. Args: file: Path to the .mv2 memory file infer: Infer schemas from stored data summary: Show schema summary Returns: Schema information or inferred schemas

List available embedding models. Shows models configured in embedder.toml. Model types: - text: Text embedding models (e.g., OpenAI text-embedding-3-large) - clip: Multimodal models for images - whisper: Audio transcription models Args: model_type: Filter by model type (optional) Returns: Li…

Encrypt a memory file. Creates an encrypted .mv2e file using AES-256-GCM encryption. The original .mv2 file is not modified. Args: file: Path to the .mv2 file to encrypt output: Output path for the encrypted file (.mv2e) password: Encryption password (required for non-interactive use) Retur…

Decrypt an encrypted memory file. Decrypts a .mv2e file back to .mv2 format. Args: file: Path to the encrypted .mv2e file output: Output path for the decrypted file (.mv2) password: Decryption password (required for non-interactive use) Returns: { "decrypted_file": string, "succes…

Run Named Entity Recognition (NER) enrichment to extract entities. Extracts entities like people, organizations, locations, dates from content. Requires NER model configuration. Args: file: Path to the .mv2 memory file all: Process all frames that haven't been enriched frame_id: Process a sp…

Memory card operations - list, stats, or filter by entity. Memory cards are structured summaries of stored content. Args: file: Path to the .mv2 memory file list: List all memory cards stats: Show memory card statistics entity: Filter cards by entity name Returns: Memory cards matching …

State management - show current memory state. Shows active session, binding status, and processing queue. Args: file: Path to the .mv2 memory file show: Display current state Returns: Current memory state information

Fact extraction - list facts or extract from frame. Facts are structured assertions extracted from content. Args: file: Path to the .mv2 memory file frame_id: Extract facts from a specific frame list: List all extracted facts Returns: List of facts with source references

Follow entity relationships in the knowledge graph. Traverses the entity graph starting from a given entity. Args: file: Path to the .mv2 memory file entity: Starting entity name link: Relationship type to follow (default: "related") hops: Number of relationship hops to traverse (default: …

Entity lookup - find information about an entity. Searches for entity mentions and related information. Args: file: Path to the .mv2 memory file query: Entity name or search query Returns: Entity information and mentions

Create a new .mv2 memory file for storing AI agent memories. The .mv2 format is a SQLite-based memory store supporting: - Full-text lexical search (Tantivy) - Vector similarity search (embeddings) - Temporal indexing Args: file: Absolute path where the .mv2 file will be created Returns: JSON …

Open and display information about a memory file. Returns metadata including: - Frame count - Index sizes (lexical, vector, temporal) - Creation date - Last modified date Args: file: Path to the .mv2 memory file Returns: JSON with file metadata and statistics

Show detailed statistics for a memory file. Returns: { "frame_count": number, "vector_count": number, "lex_index_bytes": number, "vec_index_bytes": number, "time_index_bytes": number, "total_size_bytes": number } Args: file: Path to the .mv2 memory file

Verify integrity of a memory file. Checks: - SQLite database integrity - Index consistency - Frame checksums (with deep=true) Args: file: Path to the .mv2 memory file deep: Perform thorough verification including all frame checksums (slower) Returns: JSON with verification results and any i…

Diagnose and repair a memory file. Can rebuild corrupted indexes: - Time index: chronological ordering - Lexical index: full-text search (Tantivy) - Vector index: similarity search Args: file: Path to the .mv2 memory file rebuild_time_index: Rebuild temporal index rebuild_lex_index: Rebuild …

Search in a memory file using hybrid, lexical, or vector search. Search modes: - hybrid (default): Combines lexical and vector search with RRF ranking - lex: Full-text lexical search only (Tantivy) - vec: Vector similarity search only (requires embeddings) Args: file: Absolute path to .mv2 file …

Vector-only similarity search (semantic search). Uses vector embeddings for semantic similarity matching. Requires embeddings to be generated (use memvid_put with embed=true). Args: file: Path to the .mv2 memory file query: Search query text (will be embedded) limit: Maximum results (default…

Ask a question using RAG (Retrieval-Augmented Generation). Retrieves relevant context from memory and synthesizes an answer. Requires LLM configuration in llm.toml. Modes: - hybrid: Combined lexical + semantic retrieval - lex: Lexical retrieval only - sem: Semantic retrieval only Args: file: Pa…

Show chronological timeline of frames. Returns frames ordered by creation time. Args: file: Path to the .mv2 memory file limit: Maximum entries to show reverse: Show newest first (default: oldest first) since: Filter from Unix timestamp (milliseconds) until: Filter until Unix timestamp (…

Temporal search - find when something was mentioned. Searches for content and returns results with temporal context. Args: file: Path to the .mv2 memory file query: Time-related search query limit: Maximum results (default: 10) Returns: Results with timestamps showing when topics were men…

Session management - list, start, stop, or replay sessions. Sessions track agent interactions for replay and analysis. Args: file: Path to the .mv2 memory file list: List all recorded sessions start: Start a new session with this name stop: Stop the current active session replay: Replay …

Memory binding operations - show or unbind. Bindings associate memory files with specific contexts or agents. Args: file: Path to the .mv2 memory file show: Show current binding information unbind: Remove the current binding Returns: Binding status or operation confirmation

Show system status. Displays: - Memvid version - NER model status - Embedder configuration - LLM configuration Returns: System status information

Sketch operations using SimHash for near-duplicate detection. Sketches enable fast similarity detection between frames. Args: file: Path to the .mv2 memory file build: Build sketches for all frames without them stats: Show sketch statistics Returns: Sketch build status or statistics

Nudge operations - trigger background processing. Forces processing of pending operations like: - Index updates - Embedding generation - Entity extraction Args: file: Path to the .mv2 memory file Returns: Processing trigger confirmation

Process pending operations queue. Executes any queued operations such as: - Deferred index updates - Batch embedding generation - Scheduled enrichments Args: file: Path to the .mv2 memory file Returns: { "operations_processed": number, "errors": number }

Verify integrity of a single frame. Checks: - Content checksum - Metadata integrity - Index consistency Args: file: Path to the .mv2 memory file frame_id: Frame ID to verify Returns: { "frame_id": number, "valid": boolean, "issues": [] }

Show current configuration. Displays configuration from: - embedder.toml (embedding model settings) - llm.toml (LLM settings for RAG) - Environment variables Returns: { "memvid_path": string, "config_dir": string, "embedder": {...}, "llm": {...} }

Print memvid version information. Returns: Version string (e.g., "memvid 0.1.0")

Add content to a memory file from a file or directory. Supports: text, markdown, code, PDF, images (OCR). Use embed=true for vector embeddings (semantic search). Args: file: Absolute path to .mv2 file input: Absolute path to file/directory to ingest recursive: Include subdirectories when inp…

Batch add multiple files from a directory with progress tracking. Optimized for large directories with commit batching. Args: file: Path to the .mv2 memory file input: Path to input directory recursive: Include subdirectories parallel: Process files in parallel batch_size: Number of file…

View content of a specific frame by ID. Args: file: Path to the .mv2 memory file frame_id: Frame ID (non-negative integer) raw: Show raw content without formatting/highlighting Returns: { "frame_id": number, "content": string, "uri": string, "created_at": string, "metad…

Update content of a specific frame. Replaces the entire content of the frame. For partial corrections, use memvid_correct instead. Args: file: Path to the .mv2 memory file frame_id: Frame ID to update content: New content for the frame Returns: JSON with update confirmation

Delete a specific frame from memory. This operation is destructive and cannot be undone. Args: file: Path to the .mv2 memory file frame_id: Frame ID to delete force: Skip confirmation prompt Returns: JSON with deletion confirmation

Correct/amend content of a frame. Creates a correction record preserving the original content for audit purposes. Use this instead of update when you want to maintain history. Args: file: Path to the .mv2 memory file frame_id: Frame ID to correct content: Corrected content Returns: JSON w…

Fetch content from a URL and add to memory. Supports web pages, APIs, and document URLs. Args: file: Path to the .mv2 memory file url: URL to fetch content from title: Custom title for the fetched content (optional) Returns: JSON with fetch status and created frame ID