Onboarding

An interview, not a form.
Questions that generate your files.

Your first Brainfile session is Claude interviewing you — one question at a time. The answers don't sit in a doc: they're written directly into your CLAUDE.md, seed your backlog, start your ledger, and set your Autopilot schedule. Below is the real question set from INTERVIEW-ME.md — the exact file the pack ships.

Free by email: a practical Claude Code tip every Tuesday and Thursday, plus a Sunday recap.

Start free — 30 days, no card See what it generates
The real question set

Ten short questions. One at a time.

From the shipped file (INTERVIEW-ME.md). Open your Brainfile folder in Claude Code and type start: Claude asks these one at a time, writes your answers into your files, then shows you your Brainfile. About 5 minutes; one-line answers are fine, and you can skip any.

01 · Role

"What do you do, and what's your business or project called?"

The anchor for everything else — every future session opens already knowing who it's working for.

Writes to: CLAUDE.md → "Who you are working for" + "How to work with me"
02 · Who it serves

"Who is it for? (Your customers, clients, readers, or team.)"

So Claude judges every draft against the people it's actually for.

Writes to: CLAUDE.md → "Who we serve"
03 · Projects

"What are the 1-3 things you most want to get done in the next month?"

Your month-1 projects become your first backlog entries — Autopilot pulls from these on run one.

Writes to: brain/backlog.md — fills the starter work classes
04 · Tools

"What tools, apps or sites do you work in?"

Shopify, WordPress, Google Docs, Stripe, Excel — or “just my laptop”. Loaded on session one, never re-explained.

Writes to: CLAUDE.md → "How to work with me"
05 · Recurring tasks

"What do you do every week that you'd hand off if you could?"

The chores you'd delegate become a self-refilling backlog class.

Writes to: brain/backlog.md → "Reduce one manual step"
06 · Communication

"How should I talk to you: short and to the point, or step-by-step with explanations?"

So you never have to say “shorter” twice.

Writes to: CLAUDE.md + brain/decisions.md
07 · Rules

"Name up to 3 things I must always do, or never do, for you."

These join the six built-in rules Claude reads at the start of every session.

Writes to: CLAUDE.md → "Your rules"; tracked in brain/ledger.md
08 · Good work

"How will you know I did a task well?"

This becomes the verification standard the ledger holds every task to.

Writes to: CLAUDE.md → "How to work with me"
09 · Autopilot

"Want me to work in timed runs while you're busy? If yes: how often, and how long each time?"

E.g. 1 hour every weekday. The interview also covers the honest practical bit: for runs while you're away, your computer must stay awake (Mac: caffeinate; Windows: Power & sleep).

Writes to: CLAUDE.md → "Autopilot"; brain/decisions.md
10 · End-of-run brief

"After each run, want a short written report (done / needs you / next)?"

Opt-in, and the template is yours to restructure.

Writes to: CLAUDE.md → "Autopilot"; template at templates/END-OF-RUN-BRIEF.md

Think a question is missing? Tell us what the interview should ask →

What it generates

The output, previewed honestly.

Example — generated from the sample answers shown, not a real customer

Say a solo founder answered: "I run a bookkeeping practice for e-commerce stores", month-1 goal "launch the new client-intake flow", rule #1 "never email a client without my sign-off", schedule "1h Autopilot daily, brief on." The interview writes files like these:

CLAUDE.mdexample output · labeled preview
# CLAUDE.md — Operating System

## Who you are working for
- Business: Bookkeeping practice for e-commerce stores
- Who we serve: Shopify/Amazon sellers doing $1–10M/yr

## Operating rules
1. **Never email a client without sign-off.** Draft, never send.
2. Think before act; verify before "done".
3. Every error becomes a permanent check.

## Autopilot
- Cadence: 1 hour daily (user starts the run; calendar reminder set)
- Keep-awake: caffeinate -dims during runs (macOS)
- End-of-run brief: ON → brain/runs/<date>-brief.md
brain/backlog.md · brain/ledger.mdexample output · labeled preview
## Work classes (seeded from your month-1 goals)
### 1. Ship the next client-intake improvement
- next: draft the intake form questions for new Shopify clients
- refill: after shipping, list 3 candidate improvements, keep the best.

## Ledger
| date       | directive                            | status | evidence |
|------------|--------------------------------------|--------|----------|
| 2026-07-11 | never email clients without sign-off | OPEN → rule written to CLAUDE.md | pending first verified draft |

The real thing is generated from your answers, in your project, by your Claude — nothing above is pre-written for you. Want to read the actual interview file first? It's free over MCP: point any MCP client at https://brainfile.io/mcp and call get_starter_file — no account needed.

Start free — 30 days, no card Back to the six features

Feedback on this page? Tell us →

Better together
Prompts that use your brainfile

Your brainfile gives AI persistent context — who you are, how you work, what matters right now. PromptSharp is what to do with it: a professional prompt library that runs ON that context. Brainfile is the memory; PromptSharp is the playbook. Together they compound — every prompt starts already knowing you.

A PromptSharp prompt, auto-filled from your brainfile
> Draft my weekly stakeholder update: wins, risks, and asks — in my voice, against my current priorities.
With your brainfile loaded, “my voice”, “my stakeholders” and “current priorities” fill themselves from your context files — zero re-explaining.
Browse the prompt library →

Want both? The All-Access + Brainfile annual bundle covers the pair.

Not ready to start? Stay in the loop.

A practical Claude Code tip every Tuesday and Thursday, plus a Sunday recap of the week. Free.

Or start the 30-day free trial — no card required.