Authorized, not shared
You sign in and approve the workspace — no keys pasted around, and you can revoke access in one click.
Your company's memory, in any AI you use.
MemoryCrow speaks MCP — the open standard for connecting AI to your tools. Wire it up once, and every AI on your team can remember and recall, under your own permissions.
Why MCP
One connection. Every AI remembers.
MCP is the open standard that lets an AI use outside tools. Connect MemoryCrow once and every AI your team already uses recalls from the same memory — nothing to build per tool.
Claude 
ChatGPT 
Cursor + more
One shared memory
cited · current · permission-safe
Three verbs your AI already knows.
The connection gives every AI on your team the same small, sharp set of memory actions — no syntax to learn.
remember
Hand it something worth keeping.
Your AI saves what matters as you work — a decision, a reason, a number. Keep it to yourself, or share it with your workspace or one group.
memorycrow · remember
Remember — we moved the launch to Thursday. The security review slipped.
Saved to memory
#launch shared
recall
Ask, and get the answer with its receipt.
The most relevant memories come back ranked, cited, and current as of today. When nothing genuinely fits, it says so instead of guessing.
memorycrow · recall
ASK What did we decide about refunds — and why?
Full refund within 14 days — the chargeback risk outweighed the support cost.
refunds.md · Mar '24
current
forget
Retire what's gone stale — safely.
Stop surfacing something you no longer want recalled. It shows exactly what would go before anything changes, and every forget is reversible.
memorycrow · forget · preview
FORGET the old onboarding steps
Step 3 — mail a welcome packet
Step 5 — set up the legacy VPN
Preview — nothing's changed yet. reversible
Connected in a minute.
One server URL, one sign-in. Pick your client — the shape is the same everywhere.
Claude CodeMCP-nativeOAuth 2.1· Scoped per workspace· Revocable anytime
- 1
Add the server
One command, or paste the server URL into your client config — the same shape across every major MCP client.
api.memorycrow.com/mcp - 2
Authorize your workspace
Your browser opens. Sign in, pick the workspace, review the access, and approve. Tokens refresh on their own — no keys to copy around.
OAuth · revoke anytime - 3
Just ask
Talk to your AI the way you would a teammate. It remembers and recalls for you, and every answer comes back cited.
no commands to learn
The moat
Safe by the way it's wired.
Access isn't bolted on top — it's how the connection works. The link is the permission boundary.
How permissions workScoped on every call
Recall is filtered to what you can see before an answer is ever formed — never trimmed after the fact.
Access can be pulled
Lose access in the source and you lose it here. Nothing keeps surfacing once the door closes.
Reversible by design
Forgetting previews first and can be undone. A permanent erase stays one deliberate click away.
MCP, answered. No jargon.
What teams ask before they wire MemoryCrow into their AI. Something still unclear?
Talk to usWhat is MCP, in plain terms?
MCP (the Model Context Protocol) is an open standard that lets an AI safely use outside tools. MemoryCrow speaks it, so any AI that also speaks it can store and recall your memory — no custom integration required.
Which AI tools can connect?
Any MCP-compatible client — Claude (desktop, web, and Claude Code), ChatGPT, Cursor, VS Code, and more. You connect once per client and recall from the same shared memory in all of them.
Do I need to be technical to set it up?
No. Most clients take a single command or a short config snippet, then a sign-in. Pick your client in the connect section above and copy the exact step.
Who can see what once it’s connected?
Every request runs under the access of the person who authorized it. Recall only ever returns what you’re allowed to see — both the answer and its citations — and access follows the tools you already use.
Can I disconnect or revoke access later?
Anytime. Authorization is a sign-in you can revoke, and forgetting a memory is reversible. Nothing is locked in.
Does connecting train anyone’s model on our memory?
No. Your memory is yours — we don’t train on it, and restricted knowledge stays restricted.
Give your AI a memory it can trust.
Cited, current, and only what you're allowed to see — for you, and your whole team.
A crow never forgets. Now your AI won't either.