Supermemory

Supermemory is a memory and continual learning layer for AI agents. It stores what agents learn about each user, task or tenant in a vector graph database, and brings relevant memories back into the model's context in real time. The vendor says its own model, called learner one, extracts and updates memories, merges and infers from them, and forgets outdated facts, with an understanding of time. Because memory sits outside any model provider, it works with any model and any agent harness. It is available through an API, plugins and an MCP server, and the site offers a setup prompt that can be opened in assistants such as ChatGPT, Claude and Cursor. Related offerings include SuperRAG for multimodal document extraction, a memory product that can be used as a filesystem, and connectors such as Gmail. Scale and Enterprise plans add advanced PDF extraction with OCR and descriptions of figures, while the lower plans use standard extraction. The product is aimed at developers and teams building agents and AI applications who want persistent memory without managing retrieval infrastructure. Enterprise options include SOC 2, a HIPAA business associate agreement and self hosting. Pricing is usage based with credits on every plan. Free is $0 a month with $5 of credits, Pro is $19 a month with $20 included, Max is $100 a month with $130 included, and Scale is $399 a month with $600 included. Enterprise is custom. Metered rates include $5 per million tokens for plain text memory and $10 per million for rich content.