Second Brain

Second Brain keeps project decisions, preferences and notes in one Worker in your Cloudflare account, and connected AI tools read and write them over MCP or REST. Team Edition adds a shared layer next to each member's private one. A desktop app for macOS and Windows can set everything up after a Cloudflare sign-in.

On capture, the Worker classifies a memory, checks it for duplicates and contradictions, links it to related memories and indexes it for semantic search. A separate full-text index catches exact names, ticket numbers and versions. Clients call MCP tools such as remember, recall, update, link, share and forget. Projects group memories by codebase, client or goal. Dated memories can trigger push reminders from the installed PWA. In Team Edition a memory enters the Shared layer only when its author shares it, and admins cannot read anyone's Personal workspace. Capture sources include the CLI, a Chrome extension, an Obsidian plugin, Notion, calendar and email integrations, iOS Shortcuts and session hooks for Claude Code, Codex CLI and Cursor. D1 stores memories and metadata, Vectorize holds embeddings and Workers AI runs the models. KV keeps OAuth state and cached Prompt Capsules (fixed context blocks for gateways), and static assets serve the dashboard. Five cron schedules run an hourly integration sync plus nightly and weekly jobs. If Vectorize is unavailable, capture and keyword recall keep working. Besides the desktop app, the README offers the Deploy to Cloudflare button and manual Wrangler deployment. The macOS build is signed and notarized by Apple, and Windows release builds are code-signed.

Built for: Developers, founders and small teams who use several AI assistants and want one memory store they host themselves.

What you can use it for

• Keep project decisions and preferences available in Claude, ChatGPT and Cursor
• Share a team memory while each member keeps private notes
• Capture notes from Obsidian, Notion, the browser or the CLI into one store
• Load project context when a Claude Code session starts

Deploy

• One-click · main path

  Branch main

  Deploy to Cloudflare

• CLI

  Branch main

  npm install
  npx wrangler login
  npm run db:create
  npm run vectors:create
  npx wrangler kv namespace create OAUTH_KV
  npx wrangler secret put AUTH_TOKEN
  npm run deploy

  Copy

Deploy with the button and enter AUTH_TOKEN, DIMENSION and METRIC, then add the /mcp URL to each AI client. The desktop app automates the same steps.

What it needs to run

• Workers AIAI
• D1 databasessecond-brain-db
• KV namespacesOAUTH_KV
• Vectorizesecond-brain-vectors

Good to know

Each brain supports one shared team, and the dashboard cannot yet create or switch between multiple teams. The bundled embedding models read English best, and multilingual reading requires switching models in the desktop app settings. Version 3 removed query-string token authentication, so clients must use OAuth or a bearer header. The five cron schedules use the whole Workers Free allowance of five cron triggers per account.

Cost on Cloudflare

Where this comes from

• Repositoryhttps://github.com/rahilp/second-brain-cloudflare
• App websitehttps://www.thesecondbrain.dev
• Demohttps://img.youtube.com/vi/h0JqRM0UxHE/hqdefault.jpg
• Documentationhttps://www.thesecondbrain.dev/docs
• Demohttps://www.youtube.com/watch?v=h0JqRM0UxHE

An alternative to Mem, Mem0, supermemory.