# Esment > Esment is a local-first, persistent memory system for AI assistants, built by Notas AI (https://notas.ai). It connects to Claude, ChatGPT, Cursor, Perplexity and Grok through the Model Context Protocol (MCP) and improves AI agent performance by up to 35%, while memories stay on the user's device. Esment ships as a macOS desktop app. Paid Pro, Team and Enterprise plans are available. Enterprise deployments can run fully offline: local LLM, local embeddings, local storage and a full audit trail. ## Docs - [Documentation home](https://esment.notas.ai/docs): overview of installing and using Esment - [Installation](https://esment.notas.ai/docs/install): install Esment on macOS and activate a license - [Memories](https://esment.notas.ai/docs/memories): how Esment stores, ranks and retrieves memories - [Dashboard](https://esment.notas.ai/docs/dashboard): browse, search and manage memories - [Connect assistants](https://esment.notas.ai/docs/assistants): connect AI assistants via MCP - [Claude](https://esment.notas.ai/docs/assistants/claude): Claude and Claude Code integration - [ChatGPT / OpenAI](https://esment.notas.ai/docs/assistants/openai): ChatGPT integration - [Cursor](https://esment.notas.ai/docs/assistants/cursor): Cursor editor integration - [Perplexity](https://esment.notas.ai/docs/assistants/perplexity): Perplexity integration - [Grok](https://esment.notas.ai/docs/grok): connect Grok as a custom MCP connector - [Local & public models](https://esment.notas.ai/docs/assistants/local-public): use local or public LLMs ## Product - [Home](https://esment.notas.ai/): what Esment is and why it exists - [Pricing](https://esment.notas.ai/pricing): Pro, Team and Enterprise plans - [Download](https://esment.notas.ai/download): Esment for macOS - [Enterprise](https://esment.notas.ai/enterprise): private, offline-capable deployments - [Get started](https://esment.notas.ai/start): onboarding guide ## Research Essays and research notes on AI memory from the Esment team (https://esment.notas.ai/research). Each essay is also served as plain Markdown at the same URL with a `.md` suffix. - [A Memory That Cannot Forget Is Not a Memory](https://esment.notas.ai/research/a-memory-that-cannot-forget): Agent memory research optimizes retention while auditable forgetting remains unsolved. Decay, supersession and erasure are three different operations, bi-temporal marking conflicts with the right to erasure, and provenance is the precondition for all of it. (2026-07-24; Markdown: https://esment.notas.ai/research/a-memory-that-cannot-forget.md) - [Extraction Is a Bet Against Future Questions](https://esment.notas.ai/research/extraction-is-a-bet): Every write-time transformation in an agent memory system is a prediction about which questions will be asked later. Why atomic fact extraction raises single-hop recall and lowers aggregation, and why derived layers must never replace the substrate. (2026-07-22; Markdown: https://esment.notas.ai/research/extraction-is-a-bet.md) - [The Benchmark Is Not the Territory](https://esment.notas.ai/research/the-benchmark-is-not-the-territory): Conversational memory benchmarks score question answering over a frozen corpus, which leaves the write path — supersession, deletion, provenance — unscored by construction. What the recall numbers actually measure, and what they cannot. (2026-07-21; Markdown: https://esment.notas.ai/research/the-benchmark-is-not-the-territory.md) ## Newsroom Press releases and product announcements (https://esment.notas.ai/newsroom). Each post is also served as Markdown at the same URL with a `.md` suffix. - [Esment's Agent Plugin brings persistent memory to ChatGPT, Claude and Cursor](https://esment.notas.ai/newsroom/agent-plugin-openai-claude-cursor): The Esment Agent Plugin installs a local memory server into the assistants you already use — with deterministic, always-on recall that never waits to be asked. (2026-08-06; Markdown: https://esment.notas.ai/newsroom/agent-plugin-openai-claude-cursor.md) - [Esment Cloud enters public beta](https://esment.notas.ai/newsroom/cloud-public-beta): The hosted Esment service — cloud memory for web assistants, plus bidirectional sync with the macOS app — is now live in public beta. (2026-07-21; Markdown: https://esment.notas.ai/newsroom/cloud-public-beta.md) - [Deterministic recall: memory that never waits to be asked](https://esment.notas.ai/newsroom/deterministic-recall): A session brief built from your own memories is now injected before every reply — recall stops being the model’s decision and becomes the harness’s guarantee. (2026-06-30; Markdown: https://esment.notas.ai/newsroom/deterministic-recall.md) ## Blog - [Esment Blog](https://blog.esment.notas.ai): essays, deep dives and field notes from the engineering team — persistent memory, local-first AI, agents. Its own index is at https://blog.esment.notas.ai/llms.txt and the full text at https://blog.esment.notas.ai/llms-full.txt. ## Optional - [About](https://esment.notas.ai/about): the team behind Esment - [Contact](https://esment.notas.ai/contact): talk to sales - [Privacy](https://esment.notas.ai/privacy) · [Terms](https://esment.notas.ai/terms) · [Cookies](https://esment.notas.ai/cookies) ## Full content - [llms-full.txt](https://esment.notas.ai/llms-full.txt): expanded product content plus the full text of every research essay and newsroom post in one file