The memory system for the ambitious
Give your agents a layer context to work without forgetting the rules.
Designedformulti-agentspaces
saved quietly, while you keep working
Memory is what makes you, you. Now it makes your AI, yours too.
Same question.
A different assistant.
it already knows — nothing to re-explain
Our Products
Measured, not promised
We benchmark ourselves.
Every change to the retrieval pipeline runs against standard long-conversation recall benchmarks before it ships. Internal numbers, updated as results clear review — full methodology lands with the public report on /research.
LongMemEval
500 questions · 5 memory abilities
LoCoMo
up to 35 sessions · ~300 turns each
BEAM
ICLR 2026 · conversations up to 10M tokens
ConvoMem
75k QA pairs · 6 memory categories
Two paths.
One memory engine.
Built for individuals and teams who want to get started fast, and for enterprises who need it fully on their own infrastructure.
We believe AI systems deserve persistent, structured memory — not just a context window.
NotasAI builds Esment: the local-first memory engine that gives your agents the ability to learn, recall, and adapt — exactly when they need it.
Where every AI agent remembers everything, so you never have to repeat yourself.
Frequently Asked Questions
Esment is a local-first, persistent memory system for AI assistants. It gives Claude, ChatGPT, Cursor and Perplexity a shared memory that survives across sessions, so you never have to re-explain who you are, what you are working on, or how you like things done.
Through the Model Context Protocol (MCP), the open standard for connecting tools to AI assistants. Esment runs an MCP server on your Mac; any MCP-capable client — Claude Desktop, Claude Code, Cursor, ChatGPT, Perplexity and others — connects to it in a couple of clicks. No plugins, no proprietary integrations.
Built-in memory is locked inside one chat product: what ChatGPT remembers, Cursor never sees. Esment is one portable memory layer shared by every assistant you connect — what Claude learns while you plan, Cursor can use while you code. It also gives you full control: browse, search, edit and delete every memory from the dashboard.
Esment is local-first: memories are stored on your Mac by default and nothing leaves your device. Cloud sync is optional, and enterprise deployments can run fully offline — local LLM, local embeddings, local storage — with a complete audit trail.
Yes. With persistent, relevant context available from the first message, our benchmarks measure up to a 35% improvement in agent task performance. Assistants stop wasting turns rediscovering context and start from what they already know about you and your projects.
Preferences, decisions, project context, people, conventions, and facts worth keeping — captured as you work and ranked by relevance. Esment organizes memories automatically, and the dashboard lets you curate them: pin what matters, correct what is wrong, delete what should not be there.
No. Install the macOS app, activate your license, and connect your assistants from the app. Esment handles storage, retrieval and the MCP server for you. Advanced and enterprise setups (local models, private infrastructure) are available when you need them.
Esment ships today as a signed macOS app. Because assistants connect over MCP, the memory works with whatever those assistants run on — desktop apps, editors, or the browser.
Self-service Pro and Team plans are available, plus custom Enterprise agreements with private, offline-capable deployments. A plan is what unlocks the app. See the pricing page for current plans.
Anyone who works with AI assistants daily and is tired of repeating themselves: developers using Cursor or Claude Code, teams standardizing on AI workflows, and security-minded organizations that want AI memory without sending their context to someone else’s cloud. Esment is built by Notas AI.
Long-term memory cannot be stored in accumulated plain text.
The need to process and retrieve information is key for a company.
Install Now