# Context Engine > One memory for your AI agents and your team. ChatGPT, Claude, Claude Code, Cursor and agents you build read and save to the same memory over MCP, and the people on your team use it in the app, so nobody starts from nothing. Hosted at https://app.contextengine.com; free to start. Enterprise runs it on the customer's own server instead (below). Context Engine is a memory service, not a chatbot: your AI tools do the thinking and Context Engine remembers. Memories are typed (decision, note, task, person, customer, meeting, file) and grouped into engines. An engine is a memory with a purpose, written in the owner's words, which AI tools read to decide where to save. A connection (one signed-in app or one key) reaches the engines its owner ticks, each at "Can read" or "Can read and save". Each tool gets a brief at the start of a session, searches by meaning across its engines, and every result names its engine. People use the same engines in the app: they browse every memory and who saved it, ask questions and get answers that cite their memories, read daily, weekly and monthly summaries, add files, and see the history of every change. On every plan, a whole team shares its engines at no cost per person, with roles (owner, admin, member, viewer). In the hosted app, every engine has its own database, search index and Worker on Cloudflare. Content is never used to train models. Support opens an engine only with the owner's or an admin's approval, for a stated time, and every visit is logged. Everything can be exported as Markdown and JSON. Context Engine is built and run by Relote (https://relote.com), which also helps companies put AI to work. ## Product - [Homepage](https://contextengine.com/): what it is, how it works, engines, security and privacy, FAQ - [Pricing](https://contextengine.com/pricing/): hosted plans are per engine, never per person; users are unlimited on every hosted plan, and every plan allows unlimited reads and saves. What a plan sets is how many agents (a signed-in app, a personal key or an engine key) connect to one engine, and how much it holds. Free ($0: 1 agent per engine, plus 1 per friend referred, up to 3; 1 GB per engine), Pro ($12 a month or $99 a year per engine: 10 agents, 25 GB), Unlimited ($39 a month or $299 a year per engine: unlimited agents, 100 GB), and Enterprise (below). Each hosted engine is billed on its own, monthly or yearly. - [Enterprise](https://contextengine.com/enterprise/): Context Engine on the customer's own server, single-tenant, installed and run by Relote. Priced per install, with one engine per install: From $6,000/year, updates and support included (email support, reply within one working day), the same price every year. If the customer stops paying, the installed version keeps running and only new releases and support stop. Includes roles (viewer, author, admin) and an audit log, git-versioned Markdown storage, encrypted backups, health checks and metrics, connected apps, and single sign-on (the engine is an OpenID Connect identity provider for the customer's own apps). Claude connects over MCP with OAuth. The customer brings the server, a Claude subscription and an OpenAI API key (embeddings). - [Referrals](https://contextengine.com/referrals/): share your link (in the app at https://app.contextengine.com/account/referrals); a referral counts when the friend connects at least one agent to an engine of their own. Each one adds an agent to every Free engine the referrer owns, up to 3 per engine, and adds nothing past that. The friend's first engine starts with 2 agents. Inviting somebody new to an engine counts too. - [Help](https://app.contextengine.com/help): the public product manual - [Sign in or create an account](https://app.contextengine.com/login) - [Privacy notice](https://contextengine.com/privacy/) and [terms](https://contextengine.com/terms/) ## Academy: foundations, in order - [Why Your AI Keeps Forgetting Your Business](https://contextengine.com/academy/why-your-ai-keeps-forgetting-your-business/): Every AI tool your team uses starts from zero, every time. The cause is a missing shared memory, and that can be fixed. - [What Is a Context Layer? A Plain-Language Guide](https://contextengine.com/academy/what-is-a-context-layer/): A context layer is the one memory all your AI tools read from and save to: your customers, your decisions, last week. How it works, without jargon. - [Events and Entities: How a Business Becomes Machine-Readable](https://contextengine.com/academy/events-and-entities/): Everything in your company is either something that happened or something that exists. That split is what makes a shared memory useful. - [One Brain, Many Agents: Why Shared Memory Beats Smarter Models](https://contextengine.com/academy/one-brain-many-agents/): A team of average agents with shared context beats a genius agent with amnesia. What changes when every AI tool your team uses reads one memory. - [Who Owns Your Company's AI Memory?](https://contextengine.com/academy/who-owns-your-ai-memory/): If your knowledge lives inside one vendor's chatbot, you are renting your own brain. Five tests for owning your AI memory, even on a hosted service. ## Academy: use cases and essays - [Why Your Memory Layer Should Be Allowed to Say No](https://contextengine.com/academy/why-your-memory-layer-should-be-allowed-to-say-no/): A memory that tidies itself is only safe if it proposes changes and waits for you. How that boundary works in practice, and how to check for it. - [From Action Items to Actual Tasks](https://contextengine.com/academy/action-items-to-tasks/): Every meeting produces commitments and most evaporate. Have your AI tools save each one as a task with an owner and a due date, where every tool can see it. - [Kill the Standup: Async Status From a Shared Memory](https://contextengine.com/academy/async-standup/): Your team's AI tools save status as people work. Summaries, Arrived today and Ask give you the standup picture without the meeting. - [Never Walk Into a Meeting Cold: Pre-Meeting Briefs From Shared Memory](https://contextengine.com/academy/automated-pre-meeting-briefs/): Before a meeting, ask your AI tool for a brief on who is attending, what was promised and what changed, drawn from one memory every tool shares. - [Brand Voice as Infrastructure: Make Every AI Tool Write Like You](https://contextengine.com/academy/brand-voice-as-infrastructure/): Save your voice, values and audience once as identity memories. Every AI tool your team connects reads the same ones, instead of guessing. - [Client Handovers Without the Three-Week Knowledge Transfer](https://contextengine.com/academy/client-handovers-without-knowledge-transfer/): Give each client its own engine and a handover becomes an invitation and an afternoon of reading. How agencies keep context when people move. - [Connect Your AI Tools: ChatGPT, Claude, Claude Code and Cursor on One Memory](https://contextengine.com/academy/connect-your-tools/): How AI tools connect to an engine, what each connection can read and save, and how files and your website get in. No key to paste for most. - [Stop Typing Into the CRM: Deal Memory Your AI Tools Keep](https://contextengine.com/academy/crm-that-updates-itself/): CRMs rot because nobody wants to type. Keep the deal story in one memory your AI tools save to, and let them propose CRM updates you approve. - [The Decision Log Your AI Tools Keep for You](https://contextengine.com/academy/decision-log-that-writes-itself/): Decisions made in chats and on calls vanish within weeks. Have your AI tools save each one with its reason, so "what did we decide?" always has an answer. - [Your First Engine: Giving It a Purpose Your AI Tools Can Use](https://contextengine.com/academy/give-your-engine-an-identity/): The first thing a new engine asks is what it is for. Here is why that answer steers every AI tool you connect, and how to write a good one. - [Meeting Notes That Go Somewhere: From Transcript to Team Memory](https://contextengine.com/academy/meeting-notes-to-team-memory/): Recording calls is easy. The hard part is turning transcripts into memory your whole team and every AI tool can use, instead of unread summaries. - [Every Sales Call Remembered: Building Account Memory From Call Transcripts](https://contextengine.com/academy/sales-call-account-memory/): Turn each call transcript into account memory your AI tools share, so every objection, promise and stakeholder is one question away. - [A Second Brain for the Solo Consultant](https://contextengine.com/academy/solo-consultant-second-brain/): One person, many clients, nothing dropped. An engine per client plus one for your practice, filled by the AI tools you already work in. - [Your AI Vendor Can Read Everything. Ours Has to Ask.](https://contextengine.com/academy/your-ai-vendor-can-read-everything/): Every hosted AI vendor can reach your data. What matters is what is separated by design, what is policy, and whether you can see it. Here is ours.