Eve is a portable, user-first memory layer for AI. It plugs into the AI you already use — as easily as a Google Drive MCP connector — and gives them long-horizon memory that's fully yours and fully visible. Built for every AI user: from someone who just discovered ChatGPT to a power user with a whole stack of tools.
“What the hell, Claude, I literally told you why we are not doing it that way in all caps, 30 minutes and two compactions ago.”
Anyone who's tried to keep a single AI instance going across compactions — or hold context across separate ones — knows the feeling. And its flip side: the sinking moment your favorite agent dredges up some throwaway detail about your life to over-interpret a completely unrelated thing. Memory is the holy grail inside the AI holy grail — and the systems chasing it were built for software engineers. Eve is building it for everyone: memory that just works, and that you can actually see and control.
Most memory tools are built for software engineers, wired up in code. Eve is the one designed for every kind of AI user. The same memory layer works for someone still finding their footing with ChatGPT, a knowledge worker who lives inside long-horizon projects, and a power user running a full stack of AI tools. User-first isn't a tagline here — it's who the whole product is designed around. And you don't need an enterprise plan to share: a collaborative cortex works like a supercharged Google Doc, whether that's a team's institutional knowledge or two people planning a wedding.
Eve isn't another chatbot or agent to switch to. It plugs into any AI platform the same way you'd connect a calendar — and your own model does the expensive thinking, so Eve's memory gets smarter as your model does.
Add Eve to Claude or ChatGPT like any other connector. No migration, no new habit to build.
As you talk, your agent tells you when it records a new memory, recalls an old one, consolidates, or updates — and you can ask it to, explicitly, any time.
Every memory lands in your cortex on Eve, where you can see and shape it. Value compounds the longer it runs.
The Mindmap and Memory Lane aren't decoration — they're faithful views of Eve's real memory architecture under the hood. What you see is what your agent actually navigates. Transparency isn't traded against performance here: the legibility and control are how Eve performs.
Every memory as a node, connected by edges that are thematic, social, affective, or temporal. See exactly how things relate.
Each memory is a card — gist as the title, detail inside. Edit, delete, cluster, or mark superseded with a click.
Whatever you're working on right now is binned on its own, so it never gets lost in the larger web of memories.
Deleted memories aren't gone — they're set aside and restorable at any point. Nothing disappears without your say.
Keep cortexes siloed by default, or share one across instances, tabs, and platforms — Claude planning in one tab, ChatGPT building in another, same memory.
Your agent proposes housekeeping — folding clusters under top-level nodes. Every detail is kept; a dense cortex stays fluid to navigate.
Most systems stop at a stateful stack of index cards. Eve turns the same memories your agent draws on into surfaces you can actually think with — no five other connectors required.
A Wikipedia-style synthesis of your whole cortex into chapters — every sentence footnoted back to the memory card it came from. Where you are, how you got here, and where you're headed.
A Kanban of projects you build straight from memory cards and slide from Backlog to Done — your work, organized out of what your agent already remembers.
Because the views mirror the real architecture, Eve isn't trading performance for legibility — the legibility is the architecture. We put the engine through the hardest public memory benchmarks — KAIST's MEME and LOCOMO — each scored by the benchmark's own grader, not ours. On the dependency-reasoning tasks most systems collapse on, Eve tops the field; on long-horizon recall it lands at the upper end. And every number is backed by a raw transcript you can read yourself.
Watch a real cortex form, consolidate, and get recalled across a conversation — plus a first look at Atlas and Board.
The memory that wins won't be the one with the highest benchmark score. It'll be the one you can see, control, and therefore trust — and that works for everyone, not just engineers.
Memory that's legible and yours makes your agent more capable and more trustworthy at once — capable because memory that works lets it think over long horizons, trustworthy because you can see and correct exactly what it knows. And Eve has the one thing the frontier labs can't copy without cannibalizing themselves: true cross-platform portability. Your memory shouldn't be locked to whichever model you happened to start with.
Currently in open beta — anyone can make an account. Take a look, then come build your cortex.