Get started

The point of a brain is that a second AI session, in a second conversation, can read back exactly what an earlier one wrote. This walks you there: sign up, connect your AI, write one note, and prove the second session sees it.

Before you sign up

You need an account with an AI host that can add a remote MCP connector. Claude, ChatGPT, and Codex all work; Install covers others. Whether your specific plan or workspace can add one varies by host, plan, and rollout, and changes over time, so check the host’s own current requirements rather than taking that as a promise here:

Set up your first brain

  1. Sign up. Go to cyberfoxai.com and choose Get started. Continue with Google, or enter your email and type in the 6-digit code we send you. No password, no API key.

  2. Create your first brain. A new account has no brain yet. Choose Create your first brain, give it a name, and pick a type: Project for one piece of work, Team for shared context with more than one owner over time, or Personal for your own notes. The slug fills in from the name; you can edit it.

  3. Connect your AI. Your new brain’s page shows “Connect your AI assistant” with the connector URL below. In Claude, open Settings → Connectors → Add custom connector and paste it. Using ChatGPT or Codex? Those take a few more steps, on the install page. Either way, approve the connection with the same email or Google account you just signed up with. A different one gives you a different, empty account.

    Connector URL

    https://mcp.cyberfoxai.com/mcp

  4. Write your first note. Start a new conversation and paste this. It names your brain by slug, so copy it from your own brain’s page rather than typing it from memory.

    Try this

    Use the CyberFox Brain connector. Call start_here on my brain "<your-brain-slug>", then write a note at overview.md describing what this brain is for.

    Your AI calls start_here first: that is the one mandatory first call in any session on a brain, before anything else. Then write_note.

  5. Check it in the web app. Back on your brain’s page, refresh.overview.md is listed with its path, version, and who wrote it.

  6. Open a second, separate conversation. A fresh chat, not a continuation of the one that wrote the note. Paste this.

    Read it back

    Call start_here on my brain "<your-brain-slug>", then read_note on overview.md and show me its exact content.

    You get back the same text, byte for byte, with its version number and who last changed it. That round trip, write in one session and read in a completely different one, is the entire product. Every future session, in any host, on any device, reads the same brain.

One identity everywhere

Your email (or the Google account behind it) is your identity across the web app, the connector, and every brain you belong to. Signing in to the connector with a different address gives you a different, empty account. It will not see the brain you just created.

Make CyberFox your default context

This optional, recommended instruction helps your AI remember to check your brain when a request may relate to your work. It is a pointer to the connector, not a replacement for your brain’s conventions.

Where do you use CyberFox Brain?

The instruction

I use CyberFox Brain as my source of truth for projects and durable context. When a request may relate to my work, use the CyberFox Brain connector: call list_brains when the relevant brain is unclear, then call start_here before reading or changing notes. Record durable context before finishing.

Add it to your Claude personal preferences.

This is a convenience, not a guarantee that a model will call a tool. When it does reach your brain, start_here and that brain’s BRAIN.md remain the source of its rules and conventions.

What next

Learn the habits that keep a brain useful: Daily use. Undo a mistake or see what an export contains: Recovery and Data. Installing somewhere other than Claude: Install. Something not working: Troubleshooting.