Skip to content

Platform Overview

What Engram Memory is and how it fits into your AI stack

Platform Overview

Engram Memory gives AI agents persistent memory across sessions. Store what matters, recall it when relevant, forget what you don't need, and consolidate duplicates automatically.

It ships in two forms: a self-hosted Community Edition and a hosted Cloud platform. Both share the same core operations — store, search, recall, forget, consolidate — but differ in scale, intelligence features, and deployment model.

Core Operations

OperationWhat it does
StoreSave a memory with semantic embedding and auto-classification
SearchSemantic similarity search across stored memories
RecallRetrieve relevant memories for a given context
ForgetDelete memories by ID or query match
ConsolidateFind and merge near-duplicate memories
ConnectDiscover cross-category relationships between memories

Every memory is automatically classified into one of five types: preference, fact, decision, entity, or other. See Memory Types for details.

Community Edition

A single Docker container bundles a vector database (Qdrant), a local embedding engine (FastEmbed), and the MCP recall server. No API keys, no cloud dependencies, no data leaving your machine.

docker run -d --name engram \
  -p 8585:8585 -p 6333:6333 -p 11435:11435 \
  -v engram-data:/data \
  engrammemory/engram-stack

Connect any MCP-compatible client — Claude Code, Cursor, Windsurf, or anything that speaks the Model Context Protocol — and your agent has persistent memory in minutes.

See Self-Hosted Docker Guide for full deployment instructions.

Cloud Platform

The hosted API at https://api.engrammemory.ai/v1 adds an intelligence layer on top: proprietary vector compression, automatic deduplication, multi-tier intelligent recall, webhooks, triggers, edge device fleet management, and usage analytics.

Cloud is stateless by design — it processes your request and returns enriched results. Your vectors live in your vector database, on your hardware, under your control.

See Platform vs Community for a detailed feature comparison.

How It Fits Into Agent Workflows

Engram connects to your AI agent via MCP or REST API. The agent stores memories during conversations, and relevant memories are recalled automatically in future sessions. The result: agents that learn, remember decisions, and avoid repeating mistakes.

  • MCP integration — Works with Claude Code, Claude Desktop, Cursor, Windsurf, and other MCP clients. See MCP Server.
  • SDK access — Python and JavaScript SDKs for programmatic use. See Introduction.
  • Framework adapters — LangChain, LlamaIndex, and more. See Agent Frameworks.

Next Steps