You are about to let an assistant remember you between chats. Not a sticky note in one window. A store that the next model, the next client, and the next month can all read. The useful question is not whether memory is a good idea. It is who can see it, what actually gets written, and what you can undo.
Before you connect
Keynodex Memory is a private memory server your AI talks to over MCP. It sits outside Claude, ChatGPT, Claude Code, and Codex, so the same store answers every client you connect. The product lives at https://memory.keynodex.com/. Setup is in How to Get the Most Out of Keynodex Memory.
Which AI clients work?
Claude (directory or a custom connector URL), Claude Code, ChatGPT in beta through developer mode, and Codex or any other client that supports remote MCP over HTTP. Point the client at https://memory.keynodex.com/mcp and complete Google sign-in in the browser. Per-client steps are in the setup docs.
It is not a Claude-only product. Claude is often the fastest path because of the Connectors Directory, but the store is the same no matter which client you start with.
Do I need an API key?
No. Sign-in is Google with OAuth 2.1 and PKCE, completed in the browser once per client. There is no key to generate, store, or rotate.
Privacy and trust
This is the section that should decide whether you connect.
Who can read my memories?
Your store is isolated per account. No other customer account can read it. Your assistant sees only what its own tool calls return from your own store. Keynodex staff do not browse memories as a normal part of running the product. Access for operations, abuse, or a legal demand is exceptional, logged, and limited. Details live in the privacy docs.
Is it end-to-end encrypted?
No. Content is encrypted in transit and at rest. It is not end-to-end encrypted: like most cloud products, our systems can technically access stored content in order to operate the service. If a vendor claims otherwise while still offering search, sharing, or server-side tools, treat that claim with care.
Do you train on my data?
Keynodex does not train models on your memories and does not sell your data. Memories are for your assistants to use on your behalf, not a corpus for us.
What your assistant "sees" is not a dump of your life. It is the result of specific tool calls against your store: a search, a recall, a playbook, a claim check. If you never save a fact, it is not there to retrieve.
You can delete memories from the product. For a full account erasure, email [email protected]. Ordinary questions go to [email protected] or the support page.
What actually gets stored
Three kinds of record, on purpose. Mixing them is how a store becomes noise.
- Memories are atomic durable facts. One ruling per record, still true next month.
- Playbooks are procedures that keep their own run history.
- Claims are facts stored with the command that re-checks them and a window after which they should be doubted.
It does not save your whole chat history. You choose what is remembered. You can ask the assistant to save something, or it can suggest a save and you confirm. An empty store is not a bug in the product. It is a store you have not written to yet, or you are signed in with a different Google account than the one that holds your memories. Ask for account status in chat when the store looks unfamiliar.
Permanent here means the store outlives any one model. It does not mean you cannot delete it.
Plans and limits
Free includes the full consumer tool surface. Nothing is feature-gated. The limit is 100 active memories.
What happens at the free cap?
The assistant tells you and links to billing. Nothing is deleted, and no tools stop working. Pro removes the cap and adds priority support. Plan changes happen on the website, never as a purchase inside a chat. Numbers and the upgrade path are on the plans page.
Living with it
Keynodex Memory is hosted. Improvements to existing tools apply on your next tool call. You do not reconnect for every ship.
If a brand-new tool never shows up, start a new chat first. Reconnect the connector only if it is still missing after that. Hosted products update in place; some clients cache the tool list until the next session.
If the store looks empty and you know you have saved things, check account status. The usual cause is a different Google account, not a wiped database.
Keynodex, Keynodex Memory, and PingRep
Keynodex is the company. Keynodex Memory is the MCP memory server. PingRep is the AI Representative product. They share a parent, not a login and not a memory store. Connecting Memory does not create a PingRep, and sending a PingRep does not write your Memory.
Product overview: Keynodex Memory.
Still stuck?
Setup and client steps: setup docs. What is stored and how deletion works: privacy docs. Human support: [email protected]. Full erasure and security reports: [email protected].
Common questions
Who can read my memories?
Your store is isolated per account. No other customer account can read it. Your assistant sees only what its own tool calls return from your own store. Keynodex staff do not browse memories as a normal part of running the product. Access for operations, abuse, or a legal demand is exceptional, logged, and limited. Details live in the privacy docs.
Is it end-to-end encrypted?
No. Content is encrypted in transit and at rest. It is not end-to-end encrypted: like most cloud products, our systems can technically access stored content in order to operate the service. If a vendor claims otherwise while still offering search, sharing, or server-side tools, treat that claim with care.
Do you train on my data?
Keynodex does not train models on your memories and does not sell your data. Memories are for your assistants to use on your behalf, not a corpus for us.
What happens at the free cap?
The assistant tells you and links to billing. Nothing is deleted, and no tools stop working. Pro removes the cap and adds priority support. Plan changes happen on the website, never as a purchase inside a chat. Numbers and the upgrade path are on the plans page.
Which AI clients work?
Claude (directory or a custom connector URL), Claude Code, ChatGPT in beta through developer mode, and Codex or any other client that supports remote MCP over HTTP. Point the client at https://memory.keynodex.com/mcp and complete Google sign-in in the browser. Per-client steps are in the setup docs.