Brainfile is a config bundle for Claude Code. You run it inside your own Claude Code subscription — no extra API keys, no databases, no servers. Drop it in, restart Claude, go.
Free by email: a practical Claude Code tip every Tuesday and Thursday, plus a Sunday recap.
Brainfile runs inside Anthropic's Claude Code CLI. If you don't have it yet, install it first — it's free to download (you'll need a Claude Pro or API account to use it).
That is Anthropic's native installer for Mac and Linux (no Node.js needed). On Windows, open PowerShell and paste irm https://claude.ai/install.ps1 | iex. If it says ~/.local/bin is not in your PATH, paste the line it prints, then open a new terminal window. Prefer no terminal at all? The Claude desktop app runs Claude Code too.
Full install instructions and platform-specific options (macOS, Linux, Windows): docs.claude.com/claude-code.
Brainfile installs with one command: it downloads your setup straight into your folder as plain files you own. Free: the setup interview, the safety settings and the resume skill. Basic adds the Autopilot protocol, the conversation saver and the weekly freshness feed. Pro adds the 15 role packs, under packs/. Re-run it anytime to get the newest freshness feed.
Your licence token appears on screen the moment you start your trial, and it is emailed to you, at checkout too on a paid plan. The pull command is the same for everyone; it asks you for the token.
Open a terminal, go to the project where you want Claude Code to use this OS, and run the pull command. It asks for your email and the license token from your Brainfile email; the token does not show as you paste it, so it never lands in your shell history or on a command line other programs can see:
Having Claude Code run it? Ask it to save two lines, email=you@email.com and token=YOUR_LICENSE_TOKEN, as .brainfile/licence in the folder with its file tool first; the installer reads that file (readable by you only, ignored by git).
You should now see AUTOPILOT.md, a starter CLAUDE.md, START-HERE.md, and a brain/ folder (including brain/freshness/) at the root of your project — every file plain markdown you can open and read.
bash .brainfile/update), which refuses anything unsigned, changed or older than what you have, and never changes your settings. Key fingerprint (text@brainfile.io): SHA256:IEdwg66Y11MO3strqGMmt8MBhSooEE+/Xm/ojddxN9M. The same fingerprint is in signing-keys.txt and at the top of pull.sh.Quit any running Claude Code session in this directory and start a fresh one:
Claude automatically loads CLAUDE.md and the skills in .claude/skills/. To confirm everything's wired up, type:
Claude should begin the 10-question interview. If it answers like a blank assistant instead, double-check that CLAUDE.md is in the same folder you launched claude from.
Ask Claude to read its new context and tell you what it knows. This is the fastest way to confirm the install worked end-to-end:
You should hear your own business described back, from the answers you gave in the interview, plus the six rules in CLAUDE.md. That's your signal that Brainfile is live. If it describes a blank template, type start again.
The pull delivers plain-text files you can open and read: a starter CLAUDE.md, the setup interview, the safety settings, the resume skill, the pattern files and a brain/ starter, plus the Autopilot protocol on Basic. Two files are programs, both short and readable: the conversation saver (.claude/hooks/save_conversations.py, Basic, plain Python with no network code) and the local updater (.brainfile/update, a bash script that checks Brainfile's signature before it changes anything). No binaries. The role packs below come with Pro.
.claude/rules/ (packs/README.md says how).claude/rules/ directory and Claude Code's auto-loading of CLAUDE.md are the mechanisms it relies on. Cursor uses .cursorrules, Windsurf uses .windsurfrules — different formats. The CLAUDE.md content itself is portable plain markdown you can adapt, but the agents/rules wiring is Claude Code only. We focus exclusively on Claude Code so every file leverages its full feature set instead of being a lowest-common-denominator configuration.CLAUDE.local.md, install Brainfile, then merge any project-specific rules back into the new CLAUDE.md; or (2) install Brainfile in a fresh sub-folder, diff the two CLAUDE.md files, and pull the sections you want. Every Brainfile file is plain markdown — there's no magic, just structured prose. Take what's useful, discard what isn't.brain/freshness/ and never rewrites your rules. Treat your installed Brainfile as your config and version it in git..claude/rules/. For example, a founder who also does marketing can copy .claude/rules/ files from the Marketer role pack alongside the founder pack. Claude Code loads every file in .claude/rules/ at session start.CLAUDE.md must be in the same directory you launched claude from (project root, not a sub-folder); (2) you may need to fully quit and relaunch Claude Code, not just start a new chat; (3) type start: if Claude asks the first setup question, CLAUDE.md is loaded. If all three check out and Claude is still generic, email hello@brainfile.io with your terminal output.Pick your persona, drop it in, restart Claude Code. 30-day money-back guarantee on every paid plan.
See pricing & pick your persona → Try the free builderA 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.