notion

mcp-name: io.github.n24q02m/better-notion-mcp

Markdown-first Notion for AI agents -- pages, databases, blocks, and comments in one call.

<!-- Badge Row 1: Status -->

<!-- Badge Row 2: Tech -->

<!-- BEGIN: AUTO-GENERATED-CROSS-PROMO --> <details> <summary><strong>Sister projects from n24q02m</strong> (click to expand)</summary>

</details> <!-- END: AUTO-GENERATED-CROSS-PROMO -->

Table of contents

• Features
• Install
• CLI
• Remote (HTTP mode)
• Smithery
• Status
• Documentation
• Tools
• Configuration
• Comparison
• Security
• Build from Source
• Trust Model
• License

<a href="https://glama.ai/mcp/servers/n24q02m/better-notion-mcp"> <img width="380" height="200" src="https://glama.ai/mcp/servers/n24q02m/better-notion-mcp/badge" alt="Better Notion MCP server" /> </a>

Features

• Markdown in, Markdown out -- human-readable content instead of raw JSON blocks
• 8 composite tools, 39 actions -- one call instead of chaining 2+ atomic Notion endpoints (plus config and help)
• Auto-pagination and bulk operations -- no manual cursor handling or looping
• Tiered token optimization -- ~77% reduction via compressed descriptions + on-demand help tool
• Dual transport -- local stdio (integration token) or remote HTTP (per-user or shared bearer token)

Install

Run with npx (Node.js >= 24) and a Notion integration token from <https://www.notion.so/my-integrations> (starts with ntn_):

// MCP client config (e.g. .mcp.json / Claude Code / Cursor)
{
  "mcpServers": {
    "notion": {
      "command": "npx",
      "args": ["--yes", "@n24q02m/notion@latest"],
      "env": { "NOTION_TOKEN": "ntn_your_token_here" }
    }
  }
}

The package is @n24q02m/notion (formerly @n24q02m/better-notion-mcp; old package receives one final release then freezes). The MCP Registry id stays io.github.n24q02m/better-notion-mcp, and the legacy better-notion-mcp command is kept as an alias of notion.

Or run the published Docker image (stdio):

docker run --rm -i -e NOTION_TOKEN=ntn_your_token_here n24q02m/notion:latest

Install matrix

Full per-client walkthroughs: mcp.n24q02m.com/servers/better-notion-mcp/setup/. See the Documentation section for per-client setup (Claude Code, Codex, Gemini CLI, Cursor, Windsurf) and HTTP mode.

CLI

Installing the package exposes a notion binary (plus the legacy better-notion-mcp alias; run it with npx or after a global install). With no arguments it starts the MCP server and speaks the protocol over stdin/stdout, so it is normally launched by an MCP client rather than by hand. Subcommands: auth stores the Notion integration token; config, status, logout, doctor, and token come from the shared hull CLI.

# Start the stdio server (default transport; requires NOTION_TOKEN)
NOTION_TOKEN=ntn_your_token_here npx --yes @n24q02m/notion@latest

# Start the remote HTTP server (bearer token) instead of stdio
npx --yes @n24q02m/notion@latest --http

See Configuration for the full environment-variable reference.

Remote (HTTP mode)

Deployed with the HTTP transport, the server is a remote endpoint gated by bearer-token auth -- no integration token in the server config. Point an MCP client that supports remote HTTP servers at the host you deployed it on:

// MCP client config -- remote HTTP (bearer token)
{
  "mcpServers": {
    "notion": {
      "type": "http",
      "url": "https://<your-host>/mcp",
      "headers": { "Authorization": "Bearer <your-token>" }
    }
  }
}

In multi mode (default) each user is listed in the users file (MCP_USERS_FILE, default ~/.hull/users.toml) and gets their own token and credential namespace. In token mode every client shares one token whose scrypt hash is set in MCP_AUTH_TOKEN_HASH. To stand up such an instance, see Self-Hosting (Remote Mode).

Smithery

The repo ships a smithery.yaml config for Smithery. Smithery launches the server over stdio (npx -y @n24q02m/notion) and requires no install-time config -- provide your Notion credentials at runtime through NOTION_TOKEN or the config tool (see Configuration).

Status

2026-05-02 -- Architecture stabilization update

>

Past months saw significant churn around credential handling and the daemon-bridge auto-spawn pattern. This caused multi-process races, browser tab spam, and inconsistent setup UX across plugins. The architecture is now stable: 2 clean modes (stdio + HTTP), no daemon-bridge layer, no auto-spawn from stdio.

>

Apologies for the instability period. If you encountered issues with prior versions, please update to the latest release and follow the current Setup guide -- most prior workarounds are no longer needed.

>

Related plugins from the same author: - wet -- Web search + content extraction - mnemo -- Persistent AI memory - imagine-mcp -- Image/video understanding + generation - email -- Email management - telegram -- Telegram - better-godot-mcp -- Godot Engine - crg -- Code review knowledge graph

>

All plugins share the same architecture -- install once, learn pattern transfers.

Documentation

Full docs at [mcp.n24q02m.com/servers/better-notion-mcp/](https://mcp.n24q02m.com/servers/better-notion-mcp/):

• Setup -- install methods for Claude Code, Codex, Gemini CLI, Cursor, Windsurf, mcp.json
• Modes overview -- stdio (local, integration token) and HTTP (remote, bearer token)
• Multi-user setup -- per-user credential namespace (HTTP mode)

Install with AI agent -- paste this to your AI coding agent:

Install MCP server notion following the steps at https://raw.githubusercontent.com/n24q02m/claude-plugins/main/plugins/better-notion-mcp/setup-with-agent.md

Tools

Eight composite Notion tools (39 actions) plus two infrastructure tools (config, help):

MCP Resources

Configuration

Self-Hosting (Remote Mode)

Run the HTTP server on your own host and choose an auth mode with MCP_AUTH_MODE:

• multi (default): list each user in MCP_USERS_FILE; each user's Notion token is stored in their own namespace (~/.better-notion/subs/<ns>/config.json).
• token: one shared bearer token. Mint the hash with token hash and pass it as MCP_AUTH_TOKEN_HASH.
• no-auth: no bearer check. Use only on a trusted local network.

docker run -p 8080:8080 \
  -e TRANSPORT_MODE=http \
  -e PORT=8080 \
  -e MCP_AUTH_MODE=token \
  -e MCP_AUTH_TOKEN_HASH='scrypt$...' \
  n24q02m/notion:latest

Docker Compose (loopback)

docker-compose.http.yml builds the http image target and publishes it on the host loopback interface only (127.0.0.1:8081), so the server is never reachable from other machines directly:

export MCP_AUTH_MODE=token
export MCP_AUTH_TOKEN_HASH='scrypt$...'
docker compose -f docker-compose.http.yml up

The server listens on <http://localhost:8081/mcp>. To expose it, put your own TLS reverse proxy in front of the loopback port.

Release images are published multi-arch (amd64 + arm64) to Docker Hub (n24q02m/notion) and GHCR (ghcr.io/n24q02m/notion). There is no n24q02m-hosted HTTP endpoint.

Comparison

How notion stacks up against direct competitors in each pillar:

Security

• Bearer token auth -- per-user tokens (multi) or one scrypt-hashed shared token (token)
• Rate limiting -- 120 req/min/IP on HTTP transport
• Session owner binding -- IP check + TTL for pending token binds
• Null safety -- Handles Notion API quirks (comments.list 404, undefined rich_text)

Build from Source

git clone https://github.com/n24q02m/notion.git
cd notion
bun install
bun run dev

Trust Model

Credentials are never sent to n24q02m: there is no hosted endpoint.

License

Apache-2.0 -- See LICENSE.