MyMemo AI Founder Lingfei Li Unveils Vision for Memory Agents at GenAI Week in Silicon Valley

While most generative AI tools today are focused on creating text, images, or code, one startup is aiming to help users—and their AI agents—remember instead.

That’s the thesis behind MyMemo AI, a fast-growing knowledge and memory management platform founded by Lingfei Li, an AI product designer and entrepreneur with deep experience in human-computer interaction. At Gen AI Week 2025, hosted at the Santa Clara Convention Center in Silicon Valley, MyMemo stood out with a vision that blends intelligent knowledge recall, cross-modal content management, and long-term AI collaboration—all under user control.

“We’re not just building another AI app—we’re building your second brain,” said Li in a pre-event interview.
“And not just for you, but for your AI agents and robots, too.”

At its core, MyMemo AI is a personal knowledge and memory management platform that helps users store, structure, and intelligently retrieve their digital experiences. Whether it's PDFs, websites, meeting recordings, voice notes, or screenshots, MyMemo turns them into searchable, chat-enabled memory blocks, powered by large language models (LLMs). Users can interact with their own content as if chatting with a knowledgeable assistant trained in their life.

But MyMemo isn’t just about storing data or generating summaries. Its real innovation lies in how memory is formed—not just what gets remembered.

Today, many AI tools and platforms are racing to build “long-term memory” by automatically extracting fragments from conversations. But this approach assumes everything said is worth remembering—and that’s rarely true.

“A lot of current AI systems capture memory by mining chat logs,” said Li. “But not everything we say reflects real knowledge, or even what we want to remember. Without user collaboration, you risk building a memory that’s bloated, irrelevant, or just wrong.”

Instead of blindly preserving everything, MyMemo brings the user into the loop. It allows people to filter, curate, and confirm what becomes lasting memory—ensuring that what AI remembers is not only accurate, but meaningful.

The platform allows users to build persistent memory graphs that grow as they learn, create, and work. But more importantly, it gives users control over how their memory is accessed—by whom, when, and under what conditions.

“We asked ourselves: Which memories do I want to share with my AI assistant? And further—which memories do I want my robot, my calendar, or even my doctor to remember?” Li explained.
“MyMemo gives you the power to define those boundaries. Not everything should be remembered equally. Not every AI should see the same version of you.”

Through fine-grained permission settings, users can choose to expose selected memory layers to specific AI agents—whether it’s a smart calendar assistant recalling past events, a customer support agent remembering recent issues, or a home robot adapting to lifestyle patterns.

This "agent-specific memory sharing" opens up a new model for modular, contextual AI collaboration—where AI assistants don't just access data, but truly understand a user's evolving context, within secure and user-controlled parameters.

In addition to its core software, MyMemo introduced a prototype project at Gen AI Week: RobotMemo, a humanoid AI assistant powered by the same memory infrastructure. RobotMemo can access a user’s personalized knowledge graph, hold contextual conversations, and serve as a physical memory assistant for use in home or education environments.

“Imagine a robot that knows your calendar, your favorite topics, your writing style—and helps you think more clearly,” said Li.
“We’re building that interface.”

But the ambition isn’t just about building one robot—it’s about powering many. MyMemo is designed to be an open memory platform, enabling third-party AI agents, chatbots, apps, and robotics systems to request and use user-permissioned memory in a safe and controlled way.

“Memory shouldn’t be locked inside a single device or app,” Li said.
“It should be modular, secure, and yours to share—with any AI you choose.”

Since its launch in early 2025, MyMemo AI has attracted over 100,000 users in 50+ countries, driven by traction across early adopters, students, creators, and professionals. The company operates a freemium-to-subscription model, with paying users gaining access to deeper memory graphs, agent integrations, and premium summarization tools.

And now, it’s opening its platform to developers.

MyMemo is currently building out an AI Agent SDK and Memory API, allowing other teams to embed memory-powered capabilities into their own tools.
Whether it’s a calendar assistant with historical awareness, a customer support bot with context retention, or a robot that adapts to household routines, MyMemo wants to be the memory backbone behind intelligent agents.

“We're not just building a product—we're building memory as infrastructure,” Li explained.

Lingfei Li previously led product design for intelligent systems at ByteDance and the robotics firm AgibotGo, where he focused on multimodal AI and cognitive interfaces. He holds degrees in industrial design and human-computer interaction, and has authored 16+ AI-related invention patents, many of which have been successfully commercialized.

Li founded MyMemo AI in 2024 to address what he calls “the silent crisis” of digital knowledge:

“We have access to more information than ever, but we’re forgetting faster. MyMemo is here to give people—and their AIs—long-term memory.”

Looking ahead, MyMemo plans to expand its memory-as-a-service infrastructure across consumer, enterprise, and hardware partners. The team is in early conversations with wearables manufacturers, edtech platforms, and humanoid robotics firms to embed MyMemo’s memory framework into a wide range of AI-powered applications.

“The next evolution of AI isn’t just smarter—it’s more personal,” said Li.
“Your memory should work for your agents, your tools, and your future self—and always stay under your control.”

As AI agents grow more autonomous and embedded into daily life, MyMemo is betting that the most valuable thing they’ll need isn’t another model—it’s memory.

This article was written in cooperation with Malana VanTyler