Claude, ChatGPT, Cursor and the rest read and save to the same memory, so none of them starts from nothing. Keep one for yourself, or one per client or project, with walls between them.
The problem
Claude Code forgets what you decided yesterday. Cursor does not know what ChatGPT learned this morning. You explain the same project, the same client and the same rules again, in every tool, every session.
Context Engine is the memory they share. It is a hosted service at app.contextengine.com that you sign in to with your email. You connect your AI tools once; as they work, they save what matters into it: decisions, notes, people, customers and files. When a new session starts, it starts already knowing.
How it works
Pick the app you use. ChatGPT and Claude take a few clicks; Claude Code is one command. Each connection has its own address and reaches only the engines you choose. There is no key to copy or keep safe.
Saves to Invoicing app
$ claude mcp add --scope user --transport http ce-invoicing \ https://app.contextengine.com/mcp/c/k7f2q9
Your AI tools save what matters while you work with them: a decision, a customer’s request, a person, a file. Each memory has a type, says which tool saved it, and keeps its history. Drop in a folder or a .zip of documents too.
When a tool starts, it gets a brief: what each engine is for, what happened recently and what changed since it was last here. When it needs more, it searches by meaning, and every result says which engine it came from.
Brief
Invoicing app: backend and API of my invoicing side project.
Since you were last here
Engines
Say in a sentence or two what an engine is for. Your AI tools read that purpose to decide what to save where, so the client’s decisions land in the client’s engine and your own notes stay yours.
Keeps itself in order
Context Engine writes the summaries, links what belongs together and catches duplicates, so what your tools saved last month is still useful this month.
Ask in plain words. Every answer names the memories it came from.
Card payments retry at most three times, down from five after August’s disputes. A card flagged as stolen or fraudulent is never retried.
A short account of each day, rolled into weeks and months, from what was saved.
Card retries took the day: after August’s disputes the team cut retries from five to three, and Jon drafted the runbook change. Contoso asked for invoices grouped by cost centre.
People, customers, decisions and files connect to each other, and the map shows how.
Identical ones are merged with both histories kept. Similar ones are asked about.
Every change records who made it, with which tool, and what it was before.
Works with
Sign in from the apps below, or connect anything that speaks MCP or HTTP with a key.
Made for how you work
If you build with AI
The decision you made with Claude Code on Monday is known to Cursor on Tuesday. Your conventions, your stack and why you chose it stop being something you repeat.
If you work for several clients
Give each client an engine and a connection of its own. In Claude Code, a project-scoped connection keeps each client’s repository to that client’s memory.
If your team uses AI
Invite people as owner, admin, member or viewer. See every tool that reaches the engine, and cut one off in a click. When someone leaves, what they wrote stays theirs.
Security and privacy
Built so that what you keep is yours, kept apart, and visible to you. Not hidden inside a model.
Export an engine any time as one ZIP of Markdown and JSON that opens without us. Every memory is a plain Markdown file.
Every engine gets its own database, its own search index and its own Worker. Isolation is physical, not a filter on a shared table.
Your content is sent to models only to answer you. It is not used to train anything, not sold and not used for advertising.
Support opens an engine only when you approve a request, for the time it names, and every visit is logged. A legal or safety duty is the one exception, and you are told at once.
See every tool and key that can reach an engine, with its access and when it was last used. Pause it, or cut it off.
Deleting an engine is scheduled a day ahead so you can export first. Then it is gone. Choose the EU at setup to keep its database and files there.
Single sign-on through OpenID Connect (Okta, Entra ID, Google) and optional two-step sign-in are built in. Read the privacy notice.
Questions
The rest is in the help, which is public, or write to hello@contextengine.com.
No, and it does not run your agents. Your AI tools do the thinking. Context Engine is the memory they share: they save to it as they work and read from it when they start. You can also ask it questions yourself, and every answer names the memories it came from.
ChatGPT, Claude on the web and the desktop app, Claude Code and Cursor sign in with a few clicks. Anything else that speaks MCP or HTTP connects with a key, so agents and scripts you build yourself can use the same memory.
An engine is a memory with a purpose: a few sentences, in your words, saying what it is for and what belongs in it. Your AI tools read the purpose to decide what to save there. Keep one for yourself, one per client or project, or one for a team.
Yes. Each engine is stored on its own, and each connection reaches only the engines you tick, at Can read or Can read and save. Give each client its own engine and its own connection, and nothing crosses between them.
Every engine gets its own database, its own search index and its own Worker on Cloudflare. Your content is sent to models only to answer you and is never used to train anything. Context Engine support opens an engine only when you approve a request, for the time it names, and every visit is logged.
Any time. Export gives you one ZIP of Markdown and JSON that opens without us: every memory is a Markdown file, so a text editor or a notes app reads it as it is. Deleting an engine is scheduled a day ahead so you can export first, and then it is gone.
You are told at 80%, on Home, by email and in one line of your AI tools’ brief. At the limit, Ask pauses until the month resets, while your AI tools keep reading and saving. Nothing you stored is ever deleted.