Supermemory

Founded by Dhravya Shah in 2024

Store and retrieve persistent memory for autonomous agents across different large language models

Pricing model

Free Tier, Paid

Rating

★ People love it

Time to value

Quick Setup (< 1 hour to ingest and test basic memory flows)

You can use supermemory to give AI agents persistent memory and continuous learning capabilities across different large language models. The service extracts relevant facts from user chats, files, and URLs, storing them in a dedicated vector-graph database. It automatically updates, merges, and recalls context in real time by injecting necessary tokens into an agent prompt. You can connect it via API, MCP, or developer plugins to reduce token costs, maintain cross-session context, and manage user profiles in local, cloud, or air-gapped environments.

What Supermemory does

Connect agent memory to IDEs and chat assistants using MCPStore unstructured files and URLs into a vector-graph databaseRetrieve relevant user context profiles via APIExtract text and figure descriptions from complex PDF filesDeploy memory containers inside private VPCs or local machinesInject dynamic context tokens into prompts at runtimeVector-graph database storing interconnected memory for agentsReal-time token injection into prompt contextsDocument extraction for PDFs including OCR and diagram descriptionsModel-agnostic memory architecture that works across any LLM harnessSub-400ms end-to-end recall latency on production trafficSupport for self-hosting in VPCs and air-gapped environments