Supermemory
Memory and continual-learning engine that stores agent context in a time-aware graph and injects it at runtime
Supermemory is a memory and continual learning engine for AI agents that works with any model or harness. It lets developers' agents learn about users, tasks and tenants over time, store that knowledge outside the model provider and receive it back in context at runtime.
Key features:
- Ingest files, chats, URLs and raw data into a time-aware vector-graph memory.
- Extract, merge, infer and forget facts with its learner-1 model.
- Inject relevant memories into the agent's context through hooks.
- Search and traverse stored content in one call with SuperRAG.
- Extract text from PDFs, images, audio and video.
- Connect through an API, plugins or MCP from ChatGPT, Claude, Cursor and Hermes.
Supermemory is open source and supports self-hosting, with SOC 2, HIPAA BAA and air-gapped options for larger customers. The free plan includes $5 of credits a month; Pro is $19/month, Max $100/month and Scale $399/month, each with included credits and usage billed per million tokens ingested.