supermemory

What it does

supermemory is a memory layer for AI agents. It extracts and stores context from users, tasks, and tenants in a vector-graph database, then injects relevant tokens into the model's context in real time. It is available through an API, plugins, and MCP, and works with any model or harness.

supermemory is a memory layer for AI agents that keeps learning outside the model, so it carries across models and harnesses. You send raw data such as files, chats, and URLs, and its learner-1 model updates, merges, infers, and forgets against a vector-graph database. The result is injected into your agent's context via hooks. The site reports 187ms median recall latency on production traffic and over a trillion tokens processed per month. Pricing is billed in SM tokens, the unique tokens actually ingested, with repeats and unchanged content free. The free plan includes $5 of credits renewed every month. Advanced document extraction for PDFs with OCR is limited to Scale and Enterprise plans. The site does not state whether an account is required, so that remains unknown.