Mengram
Mengram is a free, open-source AI memory API that stores and retrieves information across three distinct memory types modeled on human cognition: semantic memory (facts and preferences), episodic memo
About
Mengram is a free, open-source AI memory API that stores and retrieves information across three distinct memory types modeled on human cognition: semantic memory (facts and preferences), episodic memory (events and decisions), and procedural memory (workflows and habits). A single add() call extracts all three types automatically from a conversation, and a get_profile() call returns a ready-to-use system prompt — called a Cognitive Profile — that personalizes any LLM with everything it knows about the user. It integrates with Claude Desktop, Cursor, LangChain, CrewAI, n8n, and OpenClaw via MCP or SDK, with a free tier offering 50 memory adds and 300 searches per month.
Developers building AI assistants, coding tools, or autonomous agents that need persistent, personalized memory across sessions — particularly those who want to avoid building and maintaining their own RAG pipeline and need episodic or procedural context in addition to raw facts.
Pros & Cons
Pros
- check Three memory types in one API is a meaningful architectural difference from competitors that only store facts (semantic memory)
- check Cognitive Profile is a practical feature: one API call generates a ready-made system prompt, removing boilerplate from every integration
- check Procedural memory can self-improve — it tracks success and failure rates per workflow and auto-evolves procedures over time
- check Multi-user isolation with a single API key makes it viable for building multi-user products
- check Broad integration coverage: MCP, Python SDK, JavaScript SDK, LangChain, CrewAI, n8n, and OpenClaw plugin all supported
Cons
- close Free tier limits (50 adds/month, 300 searches/month) are low for anything beyond light personal use
- close LLM-powered reranking and smart triggers — two features that materially improve retrieval quality — require the $19/month Pro tier
- close Self-hosting is supported but adds operational overhead that contradicts the "replace your RAG pipeline" simplicity pitch
- close Extraction latency of ~2 seconds per `add()` call may be noticeable in high-frequency conversational applications
More Memory
Other tools in the same category.
