A year ago, Anthropic released the MCP standard. Paul started tinkering with it over that Thanksgiving weekend, and he’s been obsessing with Basic Memory ever since.
Today, Basic Memory is a four-person team building a knowledge platform that serves as your “ground truth.” The formula is simple: it’s all just plain text. You can read it, you can write it, you can update it, and you can delete it. Underpinning these capabilities lies the Holy Trinity of Basic Memory: ENTITIES, OBSERVATIONS, and RELATIONS. These enable the AI to understand how your data is related, what’s important, and how it all fits together across conversations.
Save your knowledge, connect it, own it forever. That’s the idea.
What Basic Memory does now
Basic Memory works across LLMs—Claude, ChatGPT, Gemini, Cursor, Claude Code. It runs locally or in the cloud. Desktop, phone, or on the web. Your notes live in plain Markdown, fully portable, always yours.
Behind the scenes, this past year we’ve moved from indexing notes into SQLite to dynamic Postgres instances per user, which means machines don’t fall over under load. We’ve switched from local storage to unlimited cloud buckets. We’ve spent hundreds of hours tuning the MCP interface so LLMs actually know how to use your knowledge effectively. When Basic Memory “just works,” it’s because we’re doing the hard work of steering the AI to find what you need.
We’ve also built migration tools (upload notes individually or in bulk, one-way sync, bidirectional sync) because getting your knowledge in should be easy. And getting it out should be even easier. Download everything as Markdown anytime you want. Your knowledge is 100% portable.
On the surface, Basic Memory doesn’t look massively different than when we launched it, apart from the themes (which honestly ARE a game changer). Under the hood, it’s a completely different machine.
How we got here
Early on, Basic Memory was a local-first memory layer for Claude Desktop. The problem was simple: Claude’s integrated memory was limited, and your notes were locked inside a corporate system. If you wanted to save conversations, you had to use awkward browser extensions or rely on AI-generated summaries that were usually garbage by the time you thought to create them.
We saw the same thing everyone else saw: AI needs better memory. A massive number of tech nerds and giant conglomerates started building solutions. We kept evaluating them and kept arriving at the same conclusion: what we’re doing is better.
Why? Because we understood something fundamental: whether you call your input an agent, a hook, a skill, or whatever else, it’s all just text. Once you see that—once you understand that all you really need is a source of truth—you see why Basic Memory works the way it does.
The grind was real. We added project support so you could organize different domains of knowledge. We made installation easier. We got it working with Claude Code (thank God). We prepared for cloud deployment and added support for multiple LLMs and frameworks.
Then we did customer interviews. More than twenty users spent their time talking with us about how they used the open-source project and what they wanted in a cloud version.
The interviews were as validating as they were illuminating. Our users wanted the same things we wanted: a system that was affordable, usable anywhere, that works across LLMs and devices. And they wanted to maintain control of their information. So that’s exactly what we built.
Moving everything to the cloud without making it suck was harder than it sounds. Our users were incredibly patient and helpful during the process. In a landscape where people are constantly chasing shiny new projects, our users have been faithful and engaged. I think that’s because they get it: they understand the value of a reliable source of truth.
We’ve also stayed committed to user privacy and control. Your data stays isolated per user. We maintain rigid boundaries in the cloud, which is its own set of challenges, but it’s non-negotiable for us.
What we’re working on
We plan to elaborate on our roadmap in an upcoming post, but in simple terms, our focus going forward is the same: more capabilities, more control, more reliability, greater simplicity. All while adding more features and depth.
If you haven’t looked at Basic Memory lately, now’s a good time. If you’re already using it, just know there’s more coming soon, and we’re excited to share it with you.
In the meantime, if you have a feature you’d like to see in Basic Memory, any feedback, a favorite recipe to share, anything at all, let us know at support@basicmemory.com. We’re always listening.
Try Basic Memory free for 7 days at basicmemory.com