Last updated: 31 August 2026
Connect your Cognitive Atoms account to Claude, and Claude can save the thoughts you explicitly ask to keep, recall your own thinking with citations, and show you what’s still open — nothing more.
The Cognitive Atoms connector links Claude to your own Cognitive Atoms account. It gives Claude exactly three abilities, and only these three:
When you say “capture this,” the thought is saved to your Cognitive Atoms account. Capture is strictly additive: the connector cannot edit, overwrite, or delete anything you have saved.
Ask what you’ve decided or captured about something, and Claude searches your own saved thinking and answers with citations to the exact thoughts it drew from. If nothing relevant is stored, it says so plainly — it never invents a memory.
Ask what’s still open, and Claude surfaces longitudinal signals from your corpus — threads you never closed, questions still unresolved — each citing the thoughts it came from.
You need a Cognitive Atoms account (email and password; the product is in beta). The connector never asks you to handle a key or token — connecting is a sign-in and a consent screen, nothing else.
Once Cognitive Atoms is listed in Claude’s connector directory, you will be able to install it directly from the listing — the sign-in and consent steps below stay the same. Until then, add it as a custom connector:
In claude.ai, open Settings → Connectors → Add custom
connector. Name it Cognitive Atoms and enter the server URL:
https://mcp.cognitiveatoms.com/mcp
Claude opens the Cognitive Atoms sign-in page. Sign in with your account email and password. No key, token, or URL ever appears for you to copy.
A consent screen shows you, in plain language, exactly what Claude is asking for — the two permissions described below, and nothing broader. Click Approve connection.
You return to claude.ai and the connector shows as connected. Try the prompts below.
The consent screen asks for two permissions, and no broader permission exists:
cognition.read — Read your saved thinking. Lets recall
and continuity search what you have captured and cite it. Strictly read-only.cognition.capture — Save new thoughts you ask to keep.
Lets capture add a thought to your account. Strictly additive: it grants no edit or
delete authority, because captured thoughts are immutable by design.Claude also asks for your go-ahead in the conversation before it saves a capture, so nothing is written without you seeing it happen.
Capture a thought:
Recall it later — in this conversation or any future one:
Claude answers from your own saved thinking, citing the exact captured thoughts it used.
See what you keep circling back to:
You can end the connection at any moment, from either side:
Either way, every token for the connection is invalidated. After you disconnect, Claude can no longer read or capture anything; if you ask it to recall, you get an honest “reconnect” state — never a silently empty account. Everything you captured stays safely in your Cognitive Atoms account, and reconnecting restores access.
The connector receives only what a tool call is explicitly given — the text you ask it to capture, or the question you ask it to recall against. It never reads Claude’s memory, chat history, summaries, or files. What you capture is stored in your own Cognitive Atoms account, used only to serve your own capture, recall, and continuity — never sold, never used to train AI models, never shared unless you explicitly share it yourself. The Privacy Policy (§4.5) covers hosted accounts and this connector in full.
Questions or trouble connecting? Email support@cognitiveatoms.com or see the support page. We aim to reply within two business days.