# Cartisien/engram-mcp

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/Cartisien/engram-mcp  
**GitHub Stars:** 1  
**npm Downloads (last month):** 202  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/cartisien-engram-mcp

## Description
Persistent semantic memory for AI agents. SQLite-backed, local-first, zero config. Semantic search via Ollama embeddings (nomic-embed-text) with keyword fallback. remember, recall, history, forget, and stats tools. Works with Claude Desktop, Cursor, and any MCP client.

## Tools
Capabilities this server exposes over MCP:

- **remember** — Store a new memory with automatic entity extraction, theme detection, and relationship mapping.
- **recall** — Retrieve relevant memories with entity, theme, and emotional context. Returns 360° relational intelligence around the query.
- **search** — Raw hybrid search with pagination and filtering. Use for precise control over results.
- **context** — Get 360° context around a specific entity, theme, or memory. Returns timeline, relationships, emotional arc, and key memories.
- **impact** — Analyze how a new piece of information would affect existing memory structure. Detects conflicts, affected entities, and themes before storing.
- **detect_changes** — Compare recent memory activity against a baseline period to identify new entities, theme shifts, emotional changes, and anomalies.
- **consolidate** — Consolidate working memories into long-term storage. Archives old memories beyond the keep threshold.
- **process_detect** — Identify recurring conversation and behavior patterns: decision loops, planning sessions, learning moments, and conflict patterns.
- **forget** — Delete memories matching criteria. Requires confirm: true. This is irreversible.
- **stats** — Get memory statistics for a session: counts by role/tier, embeddings coverage, importance scores, and graph metrics.
- **graph** — Query the entity relationship graph. Find connections, relationships, and paths between entities.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "engram-mcp": {
    "command": "npx",
    "args": ["-y","@cartisien/engram-mcp"],
    "env": {
      "ENGRAM_DB": "",
      "ENGRAM_EMBEDDING_URL": ""
    }
  }
}
```

**Requires environment variables:** `ENGRAM_DB`, `ENGRAM_EMBEDDING_URL` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation
This MCP server provides persistent semantic memory capabilities for AI agents using a local SQLite database. It supports semantic search via Ollama embeddings with a fallback to keyword search if embeddings are unavailable. The server exposes tools to remember, recall, view history, forget memories, and retrieve memory statistics. It is suitable for AI clients needing persistent, local-first memory without cloud dependencies.

**Use cases:**
- Store and retrieve agent memories persistently across sessions
- Perform semantic search on stored memories using local embeddings
- Manage conversation history and memory lifecycle
- Delete specific memories or clear memory by session or date
- Gather statistics on stored memories for analysis

**Key features:**
- SQLite-backed local memory storage
- Semantic search using Ollama nomic-embed-text embeddings
- Keyword search fallback if embeddings are unavailable
- Tools for remember, recall, history, forget, and stats
- Zero configuration and local-first operation
- Compatible with multiple MCP clients

_Summarized from the repository README — see the link above for the full text._