M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
S

Second Brain

Community
One memory layer, every AI tool.
How to install ↓ Compare GitHub source repository ↗
Category
Other #96 of 230
Stars
★ 814 Very popular
Transport
Streamable HTTP
Runtime
Hosted (no local runtime)
Credentials
API key / credential required
License
MIT
Last commit
Tools
8
57FMRS · C

This server offers rich memory features with semantic search, automatic linking, and multiple capture methods, and flexible deployment. Main limitation is that it requires a Cloudflare account and self-hosting, suited for personal use. Data control and privacy are highlights, but users take on operational responsibility.

Strongest · Documentation 15/20 Weakest · Reliability 5/20

Reliability
5/20
Security and permissions
10/20
Maintenance
14/20
Documentation
15/20
Setup experience
13/20
Why each score
Reliability 5/20
Only the README was supplied; the server manifest is not cached and no source files, CI workflows, or committed tests were available. The README declares six MCP tools (remember, append, update, recall, list_recent, forget) plus REST /capture and /health endpoints, and the happy path is coherent, but none of it is verifiable from source. Under the static calibration, reliability cannot exceed 12 without real CI and tests in the files; with zero executable evidence, a heavy deduction applies.
Security and permissions 10/20
The README discloses a self-hosted data boundary (data stays in the user's own Cloudflare account), AUTH_TOKEN sent via Authorization: Bearer, OAuth 2.0 dynamic client registration backed by KV, and a privacy statement. Install examples contain no real secrets, and no malware, credential-theft, or exfiltration vector is visible, so no red line is triggered. However, code-level credential storage, least-privilege scoping of D1/KV/Vectorize, and confirmation for the irreversible forget tool cannot be verified from the supplied material, capping the score at a mid-range 10.
Maintenance 14/20
The repository is not archived, carries an MIT license, shows an active release track (v2, v2.1, v2.2), documents a code-signing policy with release binaries built via GitHub Actions, and has Discussions enabled. Deductions: the supplied material contains no commit history, release dates, issue-response evidence, or dependency-update cadence, so sustained maintenance is plausible but not fully evidenced.
Documentation 15/20
Documentation is layered and detailed: the README covers three install paths, AUTH_TOKEN and OAuth flows, wiki links, client-specific setup, and dedicated troubleshooting sections (OAuth issues, lazy tool loading, false phishing warnings). Deductions: per-tool parameter schemas and limits are not documented, cost/limits are only summarized as 'free tier at personal scale', and no reproducible tool-call response examples (beyond /capture) are included.
Setup experience 13/20
Three install paths are clearly described (desktop app, one-click Cloudflare deploy, manual npm deploy) with client configuration examples for Claude Code, Codex, ChatGPT, and Cursor, plus an OAuth explanation. Deductions: the manual and one-click paths require several Cloudflare resource values (dimensions, metric, AUTH_TOKEN, OAUTH_KV) and post-deploy client wiring; the Windows build currently triggers a SmartScreen warning; and no CI/test evidence exists in the supplied material, so the score is capped below 15.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessConnects to a database

Best for

  • Individual users needing shared persistent memory across multiple AI tools.
  • Developers who want to self-host a memory layer on Cloudflare's free tier.
  • Users prioritizing data privacy and control, as data stays in their own cloud account.
  • Users wanting a quick setup with complete ecosystem support (desktop app, CLI, extensions, etc.).

Not for

  • Teams needing collaborative memory (currently single-user design).
  • Users wanting zero technical setup without touching Cloudflare (except desktop app).
  • Users requiring offline or local-only memory (depends on cloud services).
  • Strict compliance environments that require non-cloud infrastructure.

Required permissions

  • Requires a Cloudflare account to deploy the Worker and infrastructure.
  • Requires an AUTH_TOKEN (or password) to authorize all clients.
  • Supports OAuth 2.0 dynamic client registration; clients do not need to embed the token in configs.
  • For calendar, email, and Notion integrations, requires iCal links, app passwords, or Notion connection secrets.

Risks and side effects

  • Self-hosting means you are responsible for maintenance and updates.
  • Data is stored on Cloudflare, subject to Cloudflare's privacy policy (though within your own account).
  • Free tier may have usage limits (e.g., vector index size).
  • If you forget your password, you must log into Cloudflare to reset; you might temporarily lose access to memories.
  • Deploying with incorrect parameters (e.g., Vectorize dimensions) can degrade search functionality.

Setup

Before you start

Runtime：Hosted (no local runtime)

AUTH_TOKEN requiredsecret Password/bearer token for your Second Brain; choose a phrase or generate with openssl rand -base64 32. Shared by all clients and sent as the Authorization: Bearer header.

Install Second Brain in three ways: 1) Desktop app (recommended): Download for Mac/Windows, follow guided setup in about two minutes. 2) One-click Cloudflare deploy: Click 'Deploy to Cloudflare', enter dimensions 384, cosine metric, and your AUTH_TOKEN; after deployment, copy your Worker URL. 3) Manual deployment: Clone the repo, run npm install, npm run vectors:create, and npm run deploy. Then connect AI clients via curl scripts or manual configuration.

.cursor/mcp.json 
{
  "mcpServers": {
    "second-brain": {
      "url": "https://YOUR-WORKER-URL/mcp",
      "headers": {
        "Authorization": "Bearer YOUR-AUTH-TOKEN"
      }
    }
  }
}

Shown for Cursor. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "second-brain": {
      "url": "https://YOUR-WORKER-URL/mcp",
      "headers": {
        "Authorization": "Bearer YOUR-AUTH-TOKEN"
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add --transport http second-brain https://YOUR-WORKER-URL/mcp --header 'Authorization: Bearer YOUR-AUTH-TOKEN'

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

The client's tool list should show remember, recall, and the other memory tools; alternatively ask the agent to run recall with a natural-language query — results or a 'no memories found' reply confirms the connection.

Troubleshooting

  1. If recall returns 'no memories found' but tool call succeeds, MCP is working.
  2. OAuth login page may be falsely flagged by browsers; ignore or use another browser.
  3. In Claude Code, tools may not appear automatically; try actually invoking recall.
  4. If Cursor has OAuth registration issues, remove and re-add the MCP server.
  5. Check that Worker URL is correct and AUTH_TOKEN is set.

Things to try

Once connected, you can ask your AI assistant things like:

  • Remember: we decided the pricing model will be a monthly subscription.
  • Recall what I decided about the pricing model.
  • List my recently saved memories.
  • Link this memory to my pricing decision.

Tools 8

remember writes
Store ideas, decisions, preferences, and project context.
append writes
Add an update to an existing memory.
update writes
Replace an existing memory.
recall read-only
Find memories by meaning rather than exact wording.
list_recent read-only
Browse recently saved memories.
forget destructive
Permanently delete a memory.
link writes
Explicitly connect two memories.
connections read-only
Get direct connections of a memory.

Use cases

Save and recall project context, decisions, and preferences across AI tools.
Maintain persistent personal memory during sessions without re-explaining.
Capture information from various sources: browser, command line, Obsidian, Notion, calendar, and email.
Quickly find memories via semantic search even when wording differs.
Explore related context with memory graph and 'hops' option.

Supported clients

Claude
ChatGPT
Cursor
Codex

Listed from the project's documentation, not tested by this site.

Overview

Second Brain is a self-hosted MCP server that provides a unified persistent memory layer for AI tools like Claude, ChatGPT, Cursor, and Codex. It runs on your own Cloudflare account, using Cloudflare Workers, D1, Vectorize, and Workers AI for semantic search and memory management. Supports saving from multiple sources including CLI, browser extension, Obsidian plugin, iOS Shortcuts, Notion sync, calendar, and email capture. Data is fully under your control with OAuth authentication supported.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 659 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision ee3f6c4683ae Data synced 2026-10-19 Read the FMRS scoring method