What is shared memory for AI
Imagine you work in a team on one project: someone does research, someone talks to clients, someone writes code. And each of them does it in their own AI agent. What next? You explain it in words, send each other .md or .pdf files, retell it? A lot of details get lost. And the agents' own memory is unreliable: you open a new chat to continue working on a previous task, and something will surely have to be explained all over again.
Now imagine: you asked an agent to remember something, and it really remembers — not just the details, but everything important together with the context. And your colleague can ask their agent about it, and their agent will already know.
Shared memory for AI opens a new level of work and collaboration, increasing productivity and reducing errors and data loss.
Memory that belongs not just to one employee, but to the whole company
Remember how it used to be: documents were passed around on flash drives and by email. Then cloud technologies came into our lives, Google Drive, Notion, and that solved a lot of problems. But all these documents are for people, not for AI agents.
Right now we are living through the flash drive era for AI agents: a chat turns into an .md or .pdf and gets forwarded in a messenger, losing context and creating version confusion.
Memocore solves this problem by providing a cloud where an agent can save its knowledge.
Projects and access
You can split this cloud into projects for different tasks and different employees.
Each project is available only to those invited to it. Projects for accountants are seen only by accountants, projects for support only by support agents. You can also set the role of a specific member in a specific project: whether this employee can write to it or has read-only access.
Useful project features
Instructions. In each project you can set a single style, rules, format and language so that the agents of all employees keep memory in the same format.
History. See who changed, added or deleted what in the project.
Notifications. Get a digest or a notification for every event in the project.
Memo visibility
Sometimes something from a project needs to be shown to everyone, while the rest does not. For each memo you can set the visibility: project, company, the whole internet. If you choose "company", the agents of all employees will see this memo, while everything else in the project stays invisible to them.
Public memory
Projects can be made not only for employees of your company, but for everyone.
Private. You can add people from outside the company to these projects — you simply send them an invite by email. Such users become followers of the project and don't take up a seat in the company. They can only read this project. This is very convenient for clients: keep documentation in Memocore on the service you provide or the project you are building for a customer.
Public. Absolutely anyone can follow this project. Such a user also doesn't take up a seat in the company, and the project is read-only. Share documentation, guides, information about your company. Users can follow the project and ask their AI agent about you, and you take load off your support team.
Integrations
You can issue an access key to a project and use it in your services as a knowledge base. Build a chatbot that is easy to train through any AI agent. Even this post is just a memo in our project.
Try it yourself
Flash drives once seemed normal, until the cloud appeared. It's the same with AI agents now: forwarding chats as files seems normal until you try it differently.
Memocore works with ChatGPT, Claude and other agents that support MCP. Connect it, ask your agent to remember something. Then invite a colleague and let them ask the same question.
Publishing an MCP server to the ChatGPT and Claude directories
How to publish an MCP server to the official ChatGPT and Claude directories, using Memocore as an example: requirements, review, rejection, timelines and updates after publishing.
Memocore is now in Claude's connector directory
Memocore is now listed in Claude's connector directory. Connecting takes a couple of clicks, with no settings and no address to type in.
Dates in memos: reminders, deadlines and repeats
Memos can now have dates: one-off or repeating every day, week, month or year. Your agent sees what's coming up and reminds you through Claude routines or ChatGPT tasks.
Memo history: see exactly what your agent changed
Memo history now keeps every version: compare them line by line, see changes to text, tags and files, and remove data that got saved by mistake from the history too.
Memocore for Chrome: save web pages to your assistant's memory
The Memocore extension for Chrome: select text or take the whole page and save it to a project. Claude, ChatGPT or Cursor will find it later on their own.