MyPeople read-only pilot: this branch adds a bounded, project-scoped MCP recall service for MyPeople. It is intentionally not the upstream capture-everything configuration. See MyPeople Pilot Operations before deploying or connecting a client.
One shared memory for Claude, ChatGPT, Cursor, Codex, and every other AI tool you use.
You use Claude for some things, ChatGPT for others, and Cursor for code. But your context, including your projects, decisions, and preferences, does not move with you. You end up explaining yourself again and again.
Second Brain gives every AI tool access to the same persistent memory.
Unlike memory built into a single app, this memory belongs to you. It runs in your own Cloudflare account, stays under your control, and cannot be locked inside one AI platform.
The easiest way to get started is the desktop app. It sets everything up for you in about two minutes — no terminal, no accounts to wire together, no technical steps.
Prefer to run it yourself? Use the one-click Deploy to Cloudflare button, or follow the manual steps. See the Quick Start for all three options.
-
Memory graph. Memories now connect to each other — automatically as you save, or explicitly with the new
linkandconnectionstools. Recall can follow those connections (thehopsoption) to surface related context that a plain search would miss, and the dashboard has a new Graph tab to explore your memory visually. -
Notion sync. Connect your Notion workspace from Settings → Integrations in the dashboard. Pages you share with the connection sync into memory, stay updated as they change in Notion, and surface in recall alongside everything else. Nightly automatic sync, or on demand with Sync now.
-
Graceful degradation. If the Vectorize index is missing, recall now falls back to keyword search with a clear notice instead of failing, a new
/healthendpoint reports index status, and the dashboard shows a banner with the exact fix.
Connect Second Brain to the AI tools you already use, then save information as it comes up.
Second Brain retrieves memories by meaning rather than exact wording. Asking:
What did I decide about the pricing model?
can surface the correct memory even when the original note used completely different words.
| Tool | What it does |
|---|---|
remember |
Store ideas, decisions, preferences, and project context |
append |
Add an update to an existing memory |
update |
Replace an existing memory |
recall |
Find memories by meaning rather than exact wording |
list_recent |
Browse recently saved memories |
forget |
Permanently delete a memory |
Memory is most useful when capturing information is easy. Second Brain connects to the tools and moments where context already exists.
-
AI clients: Use
rememberdirectly within Claude, ChatGPT, Cursor, Codex, and other MCP clients. -
Command line: Run
brain remember,brain recall, and other commands from your terminal.npm install -g second-brain-cf-cli
-
Notion: Connect your Notion workspace from Settings → Integrations in the web dashboard. Create an internal connection in the Notion developer portal (a connection, not a personal access token — only connections appear in a page's Connections menu), share the pages you want remembered with it, and paste its secret — shared pages sync into memory automatically (nightly, or on demand with Sync now) and stay updated as they change in Notion.
-
Obsidian: Automatically sync notes using the Second Brain Sync plugin, also available through Obsidian Community Plugins.
-
Browser extension: Capture a page or highlighted text using the Chrome extension.
-
iPhone and iPad: Use the Brain Dump, Text Brain Dump, and Save to Brain shortcuts in
integrations/ios-shortcuts/. -
Bookmarklet: Use the lightweight bookmarklet in
integrations/bookmarklet.js.
Pick the option that fits you. They all deploy the same Second Brain into your own Cloudflare account — the difference is only how much setup you do by hand.
The lowest-friction way to get started. Download the Second Brain desktop app for Mac or Windows, open it, and it walks you through setup in about two minutes: you pick a password, sign in to (or create) a free Cloudflare account, and it builds your Second Brain in your own private space and connects your AI tools for you. After setup it becomes the app you open your dashboard with every day.
Nothing to install beyond the app itself — no terminal, no git, no configuration values to copy. Developers: see installer/ for how it works and how to build it.
The Mac build is signed and notarized by Apple. The Windows build is not yet code-signed, so Windows may show a SmartScreen "unrecognized app" notice on first launch — click More info → Run anyway. (Code signing for Windows is in progress.)
Prefer to deploy the Worker yourself without the app? Set it up in three steps.
Your AUTH_TOKEN is the password used to access your Second Brain.
Use either:
-
A memorable phrase, such as
coffee-lover-2026 -
A randomly generated token:
openssl rand -base64 32
Save this token somewhere secure. You will need it when authorizing clients and testing your deployment.
Click Deploy to Cloudflare and follow the prompts.
Enter the following values during setup:
| FIELD | VALUE |
|---|---|
| Dimensions | 384 |
| Metric | cosine |
| AUTH_TOKEN | The token you created in step 1 |
Cloudflare will provision the required resources and deploy your Worker automatically.
When deployment finishes, copy your Worker URL. It will look similar to:
https://your-worker-name.your-subdomain.workers.dev
Choose the instructions for the clients you use.
Run the command for your operating system, replacing YOUR-WORKER-URL with the Worker URL from step 2.
macOS, Linux, WSL, or Git Bash
curl -fsSL https://raw.githubusercontent.com/rahilp/second-brain-cloudflare/main/scripts/connect-ai-clients.sh | bash -s -- https://YOUR-WORKER-URLWindows PowerShell
iex "& { $(irm https://raw.githubusercontent.com/rahilp/second-brain-cloudflare/main/scripts/connect-ai-clients.ps1) } -WorkerUrl https://YOUR-WORKER-URL"The setup script configures the MCP connection and global instructions using OAuth. Your authentication token is not passed to the script.
These clients require two manual setup steps:
-
Add the provided custom instructions to the app's personalization settings.
-
Add the following URL as a custom MCP connector:
https://YOUR-WORKER-URL/mcp
Follow the client-specific instructions in the wiki for the exact menus and settings.
Your Second Brain is now ready to use across every connected client.
Replace YOUR-WORKER-URL and YOUR-TOKEN with your own values:
curl -X POST https://YOUR-WORKER-URL/capture \
-H "Authorization: Bearer YOUR-TOKEN" \
-H "Content-Type: application/json" \
-d '{"content":"second brain is working","source":"test"}'A successful response will look like:
{"ok":true,"id":"..."}How OAuth authentication works
The /mcp endpoint supports OAuth 2.0 discovery and dynamic client registration.
When you add the following URL as an MCP connector:
https://YOUR-WORKER-URL/mcp
a compatible client will:
- Detect the authentication requirement.
- Register itself with your Worker.
- Open the hosted login page in your browser.
- Ask you to enter your
AUTH_TOKEN. - Store the resulting OAuth authorization.
This means your authentication token does not need to be placed in the client configuration or included in the connector URL.
The following clients support this flow:
- ChatGPT
- Claude.ai
- Claude Code
- Codex CLI
You can also configure supported command-line clients manually:
claude mcp add --transport http second-brain https://YOUR-WORKER-URL/mcpcodex mcp add second-brain --url https://YOUR-WORKER-URL/mcpClients that cannot open a browser, such as mcp-remote in a headless environment, can use static token authentication:
Authorization: Bearer YOUR-AUTH-TOKENOAuth requires the OAUTH_KV namespace for client registrations and tokens. The Deploy to Cloudflare button provisions it automatically.
For developers who want full control from the command line. Requires Node.js and a Cloudflare account.
npm install
npm run vectors:create
npm run deploynpm run vectors:create creates the Vectorize index (384 dimensions, cosine). Wrangler then provisions the remaining Cloudflare resources automatically and fills in the required values in wrangler.jsonc. Then connect your AI clients using the same steps as Option 2, step 3.
- Setup Guide: Deploy the Worker, configure authentication, and connect AI clients
- How It Works: Semantic search, chunking, memory classification, and duplicate detection
- Connect to AI Clients: ChatGPT, Claude, Claude Code, Codex, and other MCP clients
- Capture from Anywhere: Browser extension, bookmarklet, iOS Shortcuts, and share sheet
- Web UI: Dashboard and mobile interface
- Obsidian Plugin: Installation, configuration, and sync modes
- API Reference: REST and MCP endpoints
Second Brain is built with:
- Cloudflare Workers
- D1 SQLite
- Cloudflare Vectorize
- Workers AI
- Cloudflare KV
- Model Context Protocol
- TypeScript
It runs within Cloudflare's free tier at personal scale.
Your data stays in your own Cloudflare account.
Windows builds of the Second Brain desktop app are code-signed.
Free code signing provided by SignPath.io, certificate by SignPath Foundation.
Team and roles:
| Role | Members |
|---|---|
| Authors | Rahil P (@rahilp) |
| Reviewers | Rahil P (@rahilp) |
| Approvers | Rahil P (@rahilp) |
All release binaries are built from this repository's source by GitHub Actions (installer-release.yml). Every signing request is reviewed and manually approved by an approver before a signed release is published.
Privacy statement: This program will not transfer any information to other networked systems unless specifically requested by the user or the person installing or operating it. Second Brain is self-hosted by design: during setup the desktop app talks to Cloudflare only to create resources inside your own Cloudflare account, and afterwards it communicates exclusively with your own private Second Brain. Your memories and credentials are never sent to the project maintainers or any other third party.