Mengram is an open-source AI memory API that mimics human brain architecture with semantic, episodic, and procedural memory. It enables AI applications to remember user facts, past events, and learned workflows, replacing complex RAG pipelines with a single API call for instant personalization.
Free
How to use Mengram?
Install Mengram via its Python or JavaScript SDK, or connect it to tools like Claude Desktop via MCP. Once integrated, it automatically extracts and stores three types of memory from conversations. Developers can then retrieve a unified 'Cognitive Profile'—a ready-made system prompt—to give any LLM persistent, personalized context without manual effort.
Mengram 's Core Features
Human-like memory architecture with semantic (facts), episodic (events), and procedural (workflow) memory types for comprehensive AI recall.
Generates a 'Cognitive Profile'—a unified system prompt from all memories—with one API call, enabling zero-effort AI personalization and replacing RAG pipelines.
Features autonomous 'Memory Agents' like Curator (cleans contradictions), Connector (finds patterns), and Digest (provides weekly briefs) that run automatically.
Offers unified search across all three memory types using vector, BM25, graph, and LLM re-ranking, returning relevant facts, events, and procedures simultaneously.
Includes team memory sharing, allowing teams to share context so every member's AI has access to collective knowledge via a simple invite code.
Provides smart triggers and webhooks for proactive notifications, alerting users to reminders, contradictions, or detected workflow patterns from their AI's memory.
Supports experience-driven procedural learning where workflows auto-evolve based on success/failure tracking, creating versioned, self-improving processes.
Mengram 's Use Cases
Developers building AI agents can use Mengram to give their agents persistent, contextual memory across sessions, making them more helpful and personalized over time.
DevOps engineers can implement it to learn and optimize deployment workflows from chat history, with procedures that automatically improve after failures.
Customer support teams can integrate Mengram with CrewAI to create agents that recognize returning customers and recall their entire issue history instantly.
Individuals using AI assistants like Claude or ChatGPT can connect Mengram via MCP to have their AI remember preferences, past discussions, and habits across all chats.
Product teams can leverage shared team memory to ensure every team member's AI has access to the same project context, decisions, and documented processes.
Researchers and analysts can use the AI Reflections feature to gain insights into behavioral patterns, skill clusters, and strategic observations from accumulated data.
