Quick Guide: How to Build a Claude Code “Second Brain” With Email, Notes and a Browser

An X user has shared a recipe for turning Claude Code into what they call a “second brain”: an assistant that keeps long-term notes, can read your Google mail, calendar and files, links your notes into a knowledge graph, and can browse the web. This is one person’s do-it-yourself setup built from existing tools. It is not a new Anthropic product or feature.

The author, @Bober_smart, lists seven steps and says “the whole setup process takes around 20 minutes.” That timing is the author’s claim. (Via @Bober_smart on X.)

What the setup combines

• Claude Code, Anthropic‘s coding agent, which runs in a folder on your computer. Its documentation says: “CLAUDE.md files are markdown files that give Claude persistent instructions for a project… Claude reads them at the start of every session.” Claude Code also has its own built-in auto memory.
• Google Workspace CLI (gws), a command-line tool for Gmail, Calendar and Drive. It sits in the googleworkspace organisation on GitHub, but its README states: “This is not an officially supported Google product.”
• Graphify, a third-party open-source tool. Its repository describes it as a way to “Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI”. It is aimed mainly at code; using it on personal notes is the author’s adaptation.
• Playwright MCP, Microsoft‘s tool that lets an AI agent control a web browser.
• Skills, Claude Code’s way of saving a routine. The docs say: “Create a SKILL.md file with instructions, and Claude adds it to its toolkit,” and you can run one “directly with /skill-name.”

How to set it up

1. Install Claude Code and make the folder. On macOS or Linux, Anthropic‘s install command is:

curl -fsSL https://claude.ai/install.sh | bash

Create a project folder (the author calls it SECOND BRAIN) with a subfolder named MEMORY. In the main folder, add a CLAUDE.md file with a rule such as “Save key facts from our sessions as notes in MEMORY/”. Then start Claude Code in that folder.

2. Connect Google, with as little access as possible. Install the CLI (it needs Node.js 18 or later):

npm install -g @googleworkspace/cli

On a Mac you can use brew install googleworkspace-cli instead. You also need a Google Cloud project for sign-in. If you have Google’s gcloud tool, run gws auth setup, then gws auth login. The README warns that its “recommended” preset “includes 85+ scopes”, so pick only what you need, for example:

gws auth login --scopes drive,gmail,calendar

3. Add Graphify and a browser. Graphify needs Python 3.10 or later. Install it with uv tool install graphifyy (or pipx install graphifyy; note the double “y”), then run graphify install. To keep it inside this one folder, use graphify install --project. In Claude Code, type /graphify . to build a graph; results go in a graphify-out/ folder. For browsing, add Playwright MCP:

claude mcp add playwright npx @playwright/mcp@latest

4. Wrap your routine in a skill. Create .claude/skills/<name>/SKILL.md describing what you want done, such as checking today’s calendar, scanning email and updating notes. Run it with /name.

Things to know before you start

• You are giving an AI agent your email, calendar and files. Grant the least access it needs: limited scopes, ideally a test account first. Read each command before you approve it.
• Browsing exposes the agent to prompt injection. Web pages and emails can hide instructions aimed at AI. Don’t let the agent act on your mail or Drive while it is browsing untrusted sites, and keep Claude Code’s permission prompts switched on.
• Check third-party installers before running them. gws is not an officially supported Google product, and Graphify is third-party. Graphify’s installer adds skill files and hooks to Claude Code, so review what it changes.
• Your notes are not private from the model. The MEMORY files are plain files on your computer, but anything Claude reads from them is sent to the model provider.

What this means

• A DIY recipe, not a launch. Every piece already exists; the author has put them together.
• Powerful, so keep it narrow. Start with read-only or limited access and widen it only if needed.
• The 20-minute setup is the author’s estimate. You also need to set up a Google Cloud project.

Sources

1. @Bober_smart, post on X, 14 October 2026
2. Anthropic, Claude Code docs: Set up Claude Code
3. Anthropic, Claude Code docs: Memory (CLAUDE.md)
4. Anthropic, Claude Code docs: Skills
5. Google Workspace CLI, README
6. Graphify, README
7. Microsoft, Playwright MCP, README

Related stories

• Noma Labs Details “SneakPeek”, a Now-Fixed Flaw That Let Claude Code Desktop’s Preview Pane Read Local Secrets
• AI Agent Security, 8 October (Evening): Disputed AgentCore Research, Claude Code’s Fail-Closed Hooks and Malware Aimed at AI Analysts
• Grok Agents, Part 2: Tools, Connectors and Your Own Data
• LightOn Releases ColGrep-Agent-2B, an Open Local Code-Search Agent That It Says Cuts Claude Code’s Tokens by a Third
• The OpenClaw Field Guide: How to Automate a Media Empire from Your Laptop