Your External Brain FAQ

Setup and AI assistants

What is MCP, and do I need to understand it?

MCP (Model Context Protocol) is a standard way for AI assistants to talk to outside tools. Think of it as a USB port for your AI: once it's plugged in, the assistant gains new abilities. You don't need to understand the protocol to use your External Brain. You only need to paste one URL into your assistant's settings, or run a single setup command. No coding, no JSON files.

Which AI assistants can I use?

Eight, out of the box: Claude Desktop, ChatGPT, Microsoft Copilot Studio, Gemini CLI, Codex, Claude Code, Cursor, and Windsurf. The desktop and web clients take a URL; the CLI clients use a one-line npx command. Free ChatGPT accounts can't add MCP servers yet; everything else works.

Do I need to install software, API keys, or config files?

No. For desktop clients you paste https://mcp.brain4ai.app/mcp. For CLI clients there's a one-line npx @brain-mcps/setup ... that writes the right config for you. Authentication is a magic link sent to the email you signed up with.

What changes inside my chat once it's connected?

Your assistant gains a new set of tools and starts behaving like it has a coach inside it. When you ask it to systematize a topic, it stops improvising and follows a structured methodology: asking specific questions, recording what you tell it, and producing a saved document you can return to.

Can I use my existing AI subscription?

Yes. Brain4ai isn't a separate chat product; it's something your current assistant gains access to. Bring your own Claude, ChatGPT, Gemini, etc. We never see, store, or proxy your assistant's traffic; our server only receives the methodology tool calls and the payload you choose to store.

How the methodology works

What is your External Brain, in one paragraph?

It's a guided extraction process that turns tacit knowledge, what you know but can't easily explain, into a structured, reusable document. Through indirect questions (never "tell me your method"), the assistant builds and refines a picture of how you think, what you know, and what shape the deliverable should take. At the end you walk away with a real document in the format you asked for at the start.

What does the experience look like?

You tell the assistant what you want to systematize, what your objective is, and what kind of result you want. From there the assistant carries the process through guided phases. You'll be asked indirect questions, your answers get checked and refined, and anything pulled in from the web or other tools is flagged as unverified until you weigh in. The conversation feels less like a chat and more like working with someone who already knows your field and is helping you map it out.

What do I get at the end?

Two artifacts. First, the final result itself: a standalone file in markdown, JSON, CSV, text, or HTML, whichever you specified at the start. Second, a systematization document: the audit trail of the process, including the hypotheses considered, the questions asked, and the verification status of every piece of information. The final result is what you publish or share; the systematization document is the receipt.

What's a reasoning block?

Each completed systematization is broken down into atomic, structured pieces of your expertise: a method, a principle, a decision pattern. These are reasoning blocks. They're verified, connected by how you think, and stay available to your AI across future sessions. Over time, your library of blocks becomes the external memory your assistant draws on whenever you ask it to reason about something in your field.

What if I pause halfway through?

You don't lose progress. Every systematization is saved continuously and registered as unfinished until it concludes. The next session, ask your assistant to resume; it picks up exactly where you stopped, including which pieces of information were confirmed and which were still pending.

AI behavior and safety

What stops the assistant from hallucinating bad advice?

Two things. First, the methodology constrains how the assistant moves: it can't freelance an answer at you. Second, anything from the web or external tools is recorded as unverified until you confirm it, and unverified information can't change the substance of your document; it can only flag a possible discrepancy for you to weigh in on. Your judgment stays the source of truth.

Does the assistant make decisions for me?

No. The methodology gives you clarity on what might change and why; the decision about whether to change it stays with you.

Data, privacy, and cost

Where do my results actually live?

It depends on whether your assistant calls our upload tool. If it does, the result is stored on our servers and shows up in your dashboard for viewing or downloading. If it doesn't, what happens to the result is entirely up to what you asked the assistant to do: paste it in chat, save it to disk if the assistant has file-system access, send it somewhere else, or nothing at all. You stay in control either way. The process state and saved payloads (so you can pause and resume) live in our database, but we don't store your chat transcript.

What does it cost?

There's a 15-day free trial when public access opens. During the current alpha, invited testers get free access with no trial timer running. We want feedback from real users before we lock pricing; alpha testers will get advance notice and a fair offer before anything ever moves to paid.

Alpha and support

What does "alpha" mean for me as a user?

The methodology works and the connection is stable, but rough edges are guaranteed. Some clients (especially newer ChatGPT MCP support) behave differently. Some surfaces (dashboards, billing, notifications) are still being polished. Feedback during this window has outsized influence on the final product.

How do I give feedback or report a bug?

The fastest path is inside the chat itself. Say "I have feedback about this experience" and the assistant submits it through the methodology's feedback tool: no separate form, no separate inbox. You can also reply directly to your alpha invite email; it goes to a real human.