Join waitlist
Back to products
Mengram

Mengram

AI memory API with 3 types: facts, events, and workflows

Website mengram.io
Overview

What it is

AI memory API with 3 types: semantic (facts), episodic (events), and procedural (learned workflows). One API call extracts all three automatically. Killer feature: your agent completes a task → Mengram saves the steps → next time it already knows the optimal path with success/failure tracking. Works with Claude (MCP), LangChain, CrewAI, OpenClaw. Free, open-source, Apache 2.0.

Intent

I need it when

Support multi-user AI applications with isolated memory per end-user while using a single API key

Mengram's multi-tenant architecture allows one API key to scope memories per user_id, ensuring each user gets isolated facts, events, workflows, and cognitive profiles. This enables SaaS platforms and multi-agent systems to serve many users without separate infrastructure.

Replace complex RAG pipelines with a simpler, faster memory retrieval system for LLM applications

Mengram reduces RAG setup from 15+ lines of code and multiple API keys to 3 lines with one API key. It handles chunking, embedding, and retrieval automatically, with LLM-powered reranking on Pro+ tiers and built-in synthesis via the /ask endpoint that returns citations without manual OpenAI wiring.

Enable AI agents to remember user preferences, past interactions, and learned workflows across multiple sessions

Mengram provides persistent memory with three types (semantic facts, episodic events, procedural workflows) that automatically extract and organize information from conversations. Agents can recall what worked before and improve autonomously, eliminating cold-start problems and enabling experience-driven decision-making.

Build AI coding assistants that remember developer stack, preferences, and past solutions across editor sessions

Mengram integrates natively with Claude Code, Cursor, and Windsurf via MCP server (29 tools) or SDKs. Coding assistants automatically save and recall context about the developer's environment, past fixes, and coding patterns, improving suggestion quality and reducing repetitive explanations.

Deploy AI agents that work across 23 languages with equal retrieval quality and cross-lingual search capability

Mengram uses Cohere multilingual embeddings for native support of Russian, Chinese, Spanish, Japanese, Korean, Arabic, and 17 other languages. Cross-lingual queries work natively (English query finds Russian documents), eliminating translation overhead and retrieval quality loss.

Drop

Not a fit when

  • User needs simple text storage without AI-driven memory extraction or semantic understanding
  • Application requires only basic key-value caching with no need for multi-type memory (semantic, episodic, procedural)
  • User operates exclusively in a single language and does not need multilingual memory retrieval across 23 languages
  • Project has no AI agents or LLM integrations and does not benefit from persistent agent memory across sessions
  • Organization requires on-premise deployment without custom enterprise plan negotiation
  • User needs real-time memory updates faster than <50ms latency or has extreme throughput requirements beyond Business tier limits
Commercials

Pricing

USD0 - USD99 / monthly View pricing