Mengram is an AI memory API that provides human-like memory architecture with three distinct memory types: semantic memory for facts and preferences, episodic memory for events and discussions, and procedural memory for learned workflows and processes. It automatically extracts these memory types from conversations and interactions, enabling AI systems to remember comprehensive context across sessions.

The key features include three memory types (semantic, episodic, procedural), Cognitive Profile generation for instant personalization, unified search across all memory types, knowledge graph capabilities, AI reflections for generating insights, procedure feedback with success tracking, team memory sharing, webhooks for notifications, smart triggers for proactive memory alerts, and experience-driven procedures that self-improve based on successes and failures. The system includes autonomous memory agents (Curator, Connector, Digest) that run automatically to manage memories.

Mengram works by connecting to AI tools via MCP (Model Context Protocol) or API, automatically extracting the three memory types from conversations, and providing unified access through a single API call. The Cognitive Profile feature generates ready-to-use system prompts from all memory types, eliminating the need for complex RAG pipelines.

The benefits include persistent memory across AI conversations, zero-effort personalization for AI assistants, workflow learning and optimization, team knowledge sharing, and proactive memory alerts. Use cases include AI assistants that remember user preferences and history, agent systems that learn optimal workflows, team collaboration with shared context, and automated insight generation from conversation history.

Target users include developers building AI applications, teams using AI assistants, and organizations implementing AI agents. Integrations include Claude Desktop via MCP, LangChain, CrewAI, OpenClaw, Cursor, Windsurf, ChatGPT, and Perplexity. Technical details include Python and JavaScript SDKs, cloud-based API, and open-source Apache 2.0 licensing.