A new player has entered the field of AI infrastructure. The startup Supermemory, founded full-time by Dhravya Shah, former head of developer relations at Cloudflare, recently announced a seed round of $2.6 million, aiming to solve the "memory" challenges of AI applications.
Supermemory is positioned as a general-purpose memory API for AI applications. Its core value lies in building knowledge graphs based on the data it processes and providing users with highly personalized context. It can extract and process any type of unstructured data, including files, documents, chat records, projects, emails, PDFs, and various application data streams. This multi-modal input support makes its application scenarios extremely wide, such as helping writing apps query diary entries from a month ago, or allowing video editors to retrieve resources related to specific prompts from their library.
Founder Shah emphasized: "Our core advantage is extracting insights from various unstructured data and providing more user background information to applications. Since we process multi-modal data, our solution is suitable for various AI applications, from email clients to video editors."
The chatbot and note-taking features of Supermemory allow users to add memories through text, files, or links, and can connect to mainstream applications such as Google Drive, OneDrive, or Notion. In addition, the company also provides a Chrome extension to help users easily add notes from websites. Although the company's app products are more like a "playground" for developers to understand and test the tool, its core is B2B, providing high-performance memory capabilities to other AI applications.
This seed funding round was led by Susa Ventures, Browder Capital, and SF1.vc. The investor lineup is impressive, including Cloudflare's Chief Technology Officer Dane Knecht (who provided guidance during Shah's internship), Google AI head Jeff Dean, DeepMind product manager Logan Kilpatrick, and executives from OpenAI, Meta, and Google. Joshua Browder, founder and CEO of "robot lawyer" DoNotPay, participated as an independent investor, impressed by Shah's ability to act quickly and build things. Shah had previously received an invitation from Y Combinator, but the timing was not right as investors had already joined.
In the competitive memory field, Supermemory faces startups such as Letta and Mem0 (where Shah worked briefly) that build memory layers for AI agents, as well as competitors like Memories.ai. However, Shah said that Supermemory stands out due to its lower latency. Investor Browder also affirmed Supermemory's solution, believing it offers high performance and allows applications to quickly display relevant context, meeting the growing demand for memory layers among AI companies.
Supermemory already has multiple existing customers, including the desktop assistant Cluely supported by a16z, the AI video editor Montra, the AI search Scira, the multi-MCP tool Rube from Composio, and the real estate startup Rets. It has also collaborated with a robotics company to retain visual memories captured by robots.
