Persistent, structured memory for AI Agents
🧠 𝐌𝐧𝐞𝐱𝐢𝐮𝐦 = persistent memory for LLM apps. Add one 𝐦𝐧𝐱 object and get chat history, semantic recall, and user profiles that follow users across sessions and providers. 🔄 Works with 𝐂𝐡𝐚𝐭𝐆𝐏𝐓 and 𝐂𝐥𝐚𝐮𝐝𝐞 — same memories, any model. Switch mid-conversation without losing context. ⚙️ No vector DBs or pipelines. A/B test, fail over, and route by cost — your memory layer stays consistent.
Mnexium's Records system provides CRUD operations, semantic search, and filters for schema-backed entities, complementing unstructured memory recall and enabling deterministic retrieval of business objects within agent workflows.
Mnexium's Agent State system stores task progress, pending actions, and workflow variables in a short-term, task-scoped context layer, allowing agents to resume reliably and maintain continuity across sessions.
Mnexium is model-agnostic and provider-agnostic; memories, records, and context remain accessible across OpenAI, Anthropic, and Gemini when using the same subject_id, preventing vendor lock-in and enabling flexible model selection.
Mnexium offers a drop-in API that sits between applications and models (OpenAI, Anthropic, Gemini) with 2-line code integration, handling memory assembly, chat history, records, and state without requiring custom vector databases or orchestration layers.
Mnexium provides persistent memory extraction and semantic recall that automatically learns facts, preferences, and context from conversations and injects relevant memories into future interactions, enabling personalized agent behavior without manual context management.